Mago Torres Transcript — Oct. 3, 2025

Kevin Johnson/NPF (00:00:00):

Well first I want to apologize to our speaker, Mago Torres because this week has been back and forth, back and on again, off again, and so I thank you for your flexibility on joining us today. Some of you may be familiar with Pinpoint or not. How many are using it or have used maybe one. Okay, well good. This will be new. So the aim here is to sort of explain a tool, I think a valuable tool for all of us and it goes to not only performing just traditional searches, but when you’re on dead and a volume of files lands on your desk, whether it’s the Epstein files or a volume of court records, you need to make sense of them and fast. And we’re lucky to have mago here who is a Google news trainer to introduce or reintroduce the tool that can help enrich your reporting. Mago is a longtime data journalist and a newsroom leader at the intersection of investigative journalism data and research working with journalists and news organizations to build partnerships. So please welcome Mago and I’ll turn it over to you and I’ll let you sort of set the ground rules whether you want them to ask questions as you go or do you want to reserve time for q and a later? So I’ll let you outline how you’d like to proceed.

Mago Torres/Google News Initiative (00:01:52):

Thank you Kevin. Hi everyone. I’m sure you have had an eventful week in DC when I first moved to the us, DC was the first place where I live, so I was there for two years and a half. I’m originally from Mexico City and I’m very excited to be here with all of you. So yeah, I am happy with you asking questions as we move forward. I think that as I am going to start doing the presentation, especially when we got into the demo part, is going to be like having what, understanding what kind of questions you have, it will make it richer for everyone. So yeah, feel free to move forward how you want to do it. So what I’m going to present to you today is pinpoint for journalists. I want to tell you a little bit, I’m sorry, here I want to tell you a little bit about me.

(00:02:57):

I joined the World News Initiative as a trainer for North America five months ago. Before that I was a data editor for different, I have been data editor, data researcher for different organizations. I worked for the examination doing international investigations on global public health. I was data editor also for the Latin American Center for Investigative Reporting. That is an organization based in Latin America and I have worked in different long form investigations like Panama Papers, Pandora papers, Vincent and Files and that kind of investigations and being able to have over the shelf or a tool that is helpful to work with several documents is just absolutely fantastic as someone that had to work with thousands of documents. And not only that, but also you are working with a lot of journalists and how you want to have information close and understandable for everyone and that people can share the information is key. So those are the different features that I’m going to be sharing with you today and happy to answer questions as we move forward. I want to start just to tell you what the Google News Initiative does.

(00:04:20):

We support the group work of newsrooms and journalists by offering digital tools and resources with the purpose of bolstering efforts to find, verify and tell engaging stories. We do this in from three different perspectives. One is advancing the practice of quality journalism, also strengthening and evolving publisher business models and also cultivating collaborative global news community. And we offer both in person and online sessions. And this is done in about 14 languages around the world and just a little bit of by the numbers, what the Google News Initiative has done since launching in 2018 with presence in 119 countries working with news partners in 119 countries with about 300 million in global funding. We have over almost 7,000 news partners and have been training 2.5 million journalists, which is very exciting. And just to continue doing it today, so the session today, what we’re going to do, this is our agenda is going to be an introduction to pinpoint.

(00:05:45):

I’m going to also, it’s just like what is pinpoint, how can journalists can use pinpoint share with you some use cases for you to know what has been, I’m sorry, I changed computers and I’m a little bit struggling with that. So if you see me pausing is because I’m back and forth with that and getting started with Pinpoint. So we are going to have a live demo, so by the hands that I saw, just asking again, only one person seems that have worked with Pinpoint before. That’s correct. Has anyone else heard about Pinpoint or this is your very first intro to the tool. If anyone could say

Kevin Johnson/NPF (00:06:36):

Nodding heads first time introduced to the tool.

Mago Torres/Google News Initiative (00:06:42):

This is a completely new introduction. Okay, that’s good. I’m glad. I’m happy to show you this for the first time. Yeah. Okay, so with that say, let’s start with our first topic. It’s just like introducing Pinpoint. You can go to the website, you can just Google Pinpoint and you’re going to have access to it. This is a research tool that helps reporters and academics analyze. The main purpose is to analyze large collections of documents and what we are going to see is just like what do we mean by large collections of documents and how can we talk about analyzing a lots of documents. I’m curious, have you, has any of you work with a big chunk of document for any of your stories?

Speaker 3 (00:07:44):

Yes. Yeah.

Mago Torres/Google News Initiative (00:07:46):

Okay. And what are some of the challenges that you have when you have to analyze or use different large documents for your investigations or your stories?

Speaker 4 (00:08:01):

I think typically when it’s a launch jump, right now we have a lot of launch jump coming out on Couch Oversight about the Epstein files. It can be followed when you, especially on a deadline and at least for my outlet, we want something up as quick as possible. It’s called to the pages full names and dates and without missing some key stuff at times.

Mago Torres/Google News Initiative (00:08:29):

Yeah, yeah. Okay, perfect. So we are going to see that there are many ways of approach to documents and there’s someone that has worked with lots of documents and duplicated documents and things like that. It’s just maybe some when you start using pinpoint what kind of techniques or strategies could help you just like to start and the tool offers by default just to start understanding what you have. So yeah, so you can Google and start. This tool is available for journalists and researchers if you don’t have access right now, if you have to request access, the team works very, very fast. Today is Friday but you will have access as soon as possible and if you get the access, if you have a company Google Workspace account, maybe you already have it by default. So just for you to know and if you have it, you can start also following when we do the demo. What I am doing in general pinpoint helps you with what is this helpful for? Well, for doing investigative reporting to do what was mentioned, like breaking news analysis, how you can use the tool audio and video transcript, that is one of the most common ways in how reporters and editors use the tool for fact checking, for collecting notes and searching archives.

(00:10:18):

So those are the main ways in how Pinpoint can help. So let’s go a little bit more into the details as we say. One way is to search through a thousand of documents. That is the main aspect we’re going to see. You’ll start create a collection, upload your documents and you can start doing some searches and there are all the elements that I’m going to talk about When you have audio or video files, it automatically do a transcription and create a PDF file and it also becomes searchable when you do your searches. It has also the capacity to transform when you have a set of documents that have a similar structure across the documents. You can collect information and create sortable tables or download as a spreadsheet. I don’t know if any of you have worked with data or has interest in data, but for those of us that have worked with the numbers, just distracting information, one takes a lot of time. So this is an amazing tool and can reduce time when you have a similar structure documents. That’s one of the key. And the newest thing is that the product team has been adding really amazing generative AI features inside the tool to get insights and get some information from your documents, which I think is very, very exciting.

(00:12:04):

How do you do the search through thousands of documents? As I said, you upload the document, you can upload from your Google Drive, you can upload from your computer and then the types of documents are very considerably but you start, you can search within them. It has handwriting recognition, which is amazing. Images, audio, audio transcriptions, email archives. The next slide actually, you’re going to have an overview of all the types of files that pinpoint supports and yeah, it’s very, very broad and you can use advanced search if you are familiar with advanced search elements. When you Google to be more specific about the information that you are looking for, you can look for exact match, combine keywords, exclude some terms and it’s very helpful. And the other one is understanding and filter documents by key people, organizations, states and locations. I don’t know if you’re familiar with this term, but it’s like what we will call entity extraction where we say like, oh we have a set of documents and if you work with someone that is in the data team or in the tech team in a news organization, like we need to strike the entities.

(00:13:34):

Is it like we are talking about the key names of people of organizations, of companies locations, depending on the type of documents that you’re working with.

(00:13:48):

Just to give an example, when we were working on the Panama papers, we have to go, one of the key entities that we were looking for were the name of the companies, the offshore companies that were created or maybe you are looking in some files or if you are looking into a big collection of let’s say public meetings in a very specific location and you have the list of names of people that show up in the meetings and speak and talk in these meetings and getting the names of these people, that’s an entity extraction. So these are the total types of files that you can upload to pinpoint PDF files, images files, audio files, look at all those options, emails and email archives, slideshow, word documents and also video. If for whatever reason you have a form, a type, a file type that is not supported currently at Pinpoint, you can reach out to the team.

(00:14:56):

As I mentioned at the beginning, this tool has been created for journalists and just being able to offer all the support is key for the team behind pinpoint. So if you got a very strange file through a FOIA request and you want to use it, you can ask and see what is possible for transcription and searching video and audio and video files. What Pinpoint does is you upload your file, your audio file and it’s going create a t XT file. So basically it’s going to transcribe the audio and that text is going to become searchable within, we’re going to see how you can search for terms in your collection of documents and that is also searchable. You will be able to select a part of the, if you have a quote, something that somebody said and you select that and you’re going to be able to hear it again.

(00:16:10):

This is really good. I don’t know if you noticed but I have an accent. My English is my second language and something that I have noticed sometimes is that when I speak the transcription maybe it doesn’t get everything accurate so Pinpoint offers you the option to edit the text if there needs to be a correction or a name, A name was not well transcribed or something. So you can fix that text and you can copy highlight what I mentioned. You can copy and highlight the text and something that I love is this part where it says copy highlight and highlight text and something. I don’t know how you work your draft when you work in middle in complex stories and having a draft and highlight text and when you pass it to the fact checker it’s just like I have gone through, I went through so many painful moments because of how we need to, when you are working in as a long document, you want to pass the very specific page and quote from a document, like a very specific sentence and it’s going to pass through an editor, through a collaborator, through a fact checker and just being able to provide that very specific quote, just remember this, the copy highlight is just like, it’s a very small thing that pinpoint of pinpoint offers but I think is when you have to take a lot of care of not missing a piece of information I think is an amazing tool and I’m going to show you how to do that.

(00:17:52):

And the other thing I was mention is how it can transform this similar structure information or documents into searchable tables and then you can work as a spreadsheet and do some form of data analysis and I think it’s just very, very helpful. And the generative AI features I’m going to pass, I have a list of all these links. I am going to send them to Kevin and I think Kevin and can provide all of these to you for having access to the generative AI features you have for early access, you have to opt in, they’re not going to show up by default, so if you already went to the website, you’re going to see that apply for early access to generative ai. If you want to do it, you just have to opt in and because it’s in what happened? Oh yeah, we’re here. Yeah. And generative AI features is vary by country but everyone is in the us I think so, yeah. But I’m going to also provide the link. So if you are curious, if you have to travel, if there are changes or if you work with newsroom or editors, there are somewhere else just and you want to review that, I’m going to provide those links for you to have ’em. How can the generative eye features help? You can ask is to ask questions to understand better your collection of documents.

(00:19:44):

I dunno if you have used, I was speaking Spanglish, some type of chatbot or you have used Gemini, create your prompt and use and also the chatbot is going to help you through the content. We’re going to say, well these documents talk about this. Maybe you want to ask about this or ask about that. So it offers some insights there. We’re going to see how that works. Something that it’s very relevant in for us journalists is to know to take care of the privacy of the documents that we are working with. So we just always communicate documents, collections, everything is private by default. You can make collections public but you have to go through a whole selection so you won’t make your collection of documents public by mistake and it really has steps there and is protected by Google industry leading privacy and security technologies. How are we going so far? Is there any questions so far?

Michael Williams | CNN (00:21:03):

This is Michael from CNN. How does Google respond to government requests for information or subpoenas with those documents? It makes me very nervous uploading all my reporting material and where can it be accessed like that?

Mago Torres/Google News Initiative (00:21:17):

Yeah, Google receive when receive requests. All the Google notifies users very, very quickly. But what I would say is if there’s any materials that you don’t feel comfortable to put on your Google Drive or put on an email treat pinpoint the same way. I have worked in documents where it’s just like you don’t put them anywhere. So just being mindful of that. I’m going to include in my list of links and thank you for asking that. I’m going to put in this the Google publishes the transparency of the request, so there’s a transparency report related to this. Yeah. Okay. Any other question? Nope. Okay. I’m going to show you some use cases. We didn’t test this but let’s see. Can you just react somehow if you can hear the audio of this video that I’m going to put? Yeah, it’s two minutes.

Speaker 6 (00:22:38):

Okay. What is so critical about local journalism is building trust. You have to build a relationship with people in the community so that they are willing to be open with you and that takes time.

Speaker 7 (00:23:14):

Pinpoint is an AI powered tool for sorting through large sets of documents and pulling out specific information that you’re looking for. We saw that Hillsborough County had the highest rate of lead poisoning in all of Florida. We acquired thousands of pages of documents covering all sorts of things that were going on inside of this factory. We could see that the air lead level was above the regulatory standard. The story won the Pulitzer. As a result of our reporting, the regulators changed their practices and it stopped there.

Speaker 8 (00:23:50):

We are able to do a lot more as a small team, we produce over 25 different podcasts so anything that cuts down the production time is very helpful. Now with pinpoint, when we upload the audio, you get the transcription minutes and it’s very accurate. You have a dedicated space where all your transcriptions are and you can tag them and it becomes really easy for you to cross-reference.

Speaker 6 (00:24:16):

There were these plans to convert a hotel into a shelter for the in-house to Asheville and those plans had somehow been waylaid. There was just a lot more complexity to the story than we ever imagined. The outset being able to connect the dots from those documents, I don’t think we would’ve been able to do that without pinpoint

Speaker 8 (00:25:04):

All this saves time and helps in reaching out to more people, addressing more issues.

Speaker 6 (00:25:09):

Democracy doesn’t just exist in the White House or in Congress. It exists here in the city council meetings and I’m excited about the possibility, especially on a small team to carve out the time and space to do these deeper investigations. What is causing the

Speaker 7 (00:25:24):

Delay? I think that local reporters who can embrace AI tools and technology are going to find that they have more time to spend on the human elements of their stories and the stuff that only they can do.

Mago Torres/Google News Initiative (00:25:48):

That is an overview of some of the cases reporters, newsrooms in different parts of the world that have used pinpoint. Other cases is the Boston Glove. The main feature that used was search and OCR that is transforming PDFs, images or whatever in text that is searchable and they won the Pulitzer Prize in this investigation that in some parts use pinpoint. Maria Reza has been a user since the beginning. Blacklight was its first name before it was pinpoint and also did a very important investigation that helped to search through historical documents. I think he doesn’t work there anymore, but this is in Argentina. Argentina and it has pinpoint for transcription and has helpful, was very helpful in breaking news, breaking breaking news stories or they’re just extracting the text very quickly. Settle. This is an investigation in Denmark. They use transcription and search. They had to do a lot of handwritten letters from early 19 hundreds so Pinpoint was very relevant for them to be able to search through handwritten and that also came OCR and struck the information and they won the most, I think it’s the highest investigative journalism award in Denmark and also radio free Europe use search and OCR R saying to be able to get texts from one handwritten notes and they have a collection of documents public I think.

(00:28:01):

Yeah. Okay, so these are some of the cases. So I’m going to show you very quickly, I’m curious how many of you are looking into pinpoint in your devices or are you following trying to follow with pinpoint have you had access? Cool. That sounds great. I’m going to just show you a little bit. I’m going to be alternating between my presentation and start doing a live demo for you and we can see what we find. Sounds good.

Sydney Clark/NPF (00:28:32):

Mar, I have a quick question.

Mago Torres/Google News Initiative (00:28:34):

Yes, this is

Sydney Clark/NPF (00:28:35):

Sydney Clark with National Press Foundation for the video files. Are there any file size and limitations?

Mago Torres/Google News Initiative (00:28:44):

Yes, yes, there are size limitations and actually sometimes what I will say is I don’t have the number right now. I’ll look for it, but if you have a very large size and pinpoint is not letting you upload it, you can reach out to the pinpoint pinpointing and they can help you if there’s a need of an extended capacity for your documents. Yeah, what is going to happen with very large text files, for example, if you have a 500 pages document is that it splits it in different documents but if you have some specific needs because of if you have some very specific needs you can out and say like I need this, how can I solve this issue? Because it is not that very common situation that all the documents are super long or all the video files or audio files are super heavy. Thank you. Is this helpful?

Sydney Clark/NPF (00:29:57):

Yes, thank you.

Mago Torres/Google News Initiative (00:29:58):

Fantastic. Okay. Okay, so are you looking? Yeah, when you start, this is what you’re going to see as is what I was mentioning when you want to have early access to the generative AI features, this is where you have to apply to fill the form and here there are also cases, studies, other cases studies. I mentioned some of them, some are in the video that we saw, but you can also, if you want to have a little bit of inspiration or just know what other people are doing, that’s, that’s a good place to start. When you go to get started, what you’re going to see is this is how it looks like for me because I already have some collections. I already have collections here, but you can start a new private collection. You can start, put the name of the collection. I’m just going to put October three and it’s going to start, it’s going to look empty but I can upload the documents from here, either connective to my drive or my computer and once you upload your documents it’s gonnas going to start processing processing and they’re all available. It is going to tell you and it’s going to look something like this. All your documents are going to be in the same place. They’re going to look like this. I want you to see some details. These very specific collection that you are seeing in this screen has some documents were images. So you can see here says JPG and it says PDF. So this was a transformation of the document. So that’s how they’re going to appear. I think there’s another one they’re here.

(00:32:16):

These are some of the elements if you go to explore, this was my workspace.

Speaker 3 (00:32:52):

There you go,

Mago Torres/Google News Initiative (00:33:01):

There you go. You are going to see collections of documents that other organizations have made public here where it says all organizations, you can see all the newsrooms that have published these stories that have public collections and you can search by terms. I would recommend to start searching here. Maybe you want, there could be documents that can be helpful for you and you can look there. I maybe want to look for, there are some documents from CNNI see someone mentioned that is from CNN. The Washington Post has public collections. I want to see AP and I can see some of the collections that they have. Maybe there’s something here that I want to see sign or on or after this collection of documents if I want to see something. So I think it’s very recommended to look into just public collections. Maybe there’s something that can be helpful and let’s say that I want to see these documents. I’m very sorry if I go to one document, the AP has this one and I can go here and I can download the original file. Once you make the collection public you can do this. So if you want to reuse it, that’s something that you can do because it’s already public.

(00:34:42):

If you can see here this collection that has some tabs, that is one of the features when you create your collection. I’m very sorry, my computer is very, very slow and it’s having some issues so I’m going to try to solve this situation. Let’s see.

Speaker 3 (00:35:06):

I’m going to close this. Okay,

Mago Torres/Google News Initiative (00:35:20):

Once I went back to this NASA demo collection is a collection that we have. This is from NASA files. I uploaded all my documents and by default this is telling me you have 40 documents here. If every time I see this store, I know it’s a AI feature. So let’s start with the not AI features in this collection and I will go to these three lines. And this is what I was saying about entity extraction, right? Immediately by default pinpoint is telling me the most common names and in how many documents this name appear, where are the most common organizations and how many documents this organization and locations. So let’s say that I want to know, I only want to see those that mentioned Voice of America and if I click here, it’s going to filter them out.

(00:36:21):

I can remove that filter here. Maybe I want these two files to have a specific tack because I want to be start organizing my own information, information so I can select them, start a tag, create a new tag and write Voice of America. I’m going to make it orange so I now have a tag. So next time I want to remove it here and I am going to be able to organize this information here. Same with the names. All the names that appear here and this is what I was saying, like entity extraction, everything is displayed in English. What happens maybe I have a collection of documents that are in Spanish, so I want to see the original name. Let’s say I’m just making an assumption, I’m going to change it to Spanish and America, Latin America and it’s going to change some of the names of names. For example, instead of institute is going to say, so you can select the language of your files and your either text or audio. Just notice that this applies for the whole collection. I’m going to go back to English. Audio files is a little bit more limited in the languages available but you can see everything that is available and it went back to English when you want.

Kevin Johnson/NPF (00:38:17):

I have a question

Mago Torres/Google News Initiative (00:38:18):

Kevin. Yes,

Kevin Johnson/NPF (00:38:20):

I see the filter collection there. Is that something that needs to be set or is that set a automatically? Do

Mago Torres/Google News Initiative (00:38:27):

You need to set it how you do that? Yeah, I’ll go again. Let’s say that now I want to identify, I’m very interested in Boss Alre and that’s the person that I want to see. So I selected the documents or I selected all of them, but I can just select them all here and here it appears this little icon that says labeled documents and I can create it. I already have something called the tag, so I’m just going to highlight it. But if I want to create a new one, I will create it here and I apply and now I have it and it appears here as a filter. Can you see on top below the search space? So if I remove it from there it removes and they appear visually here and here Pinpoint is also reminding me that labels that already exist.

Speaker 3 (00:39:34):

Okay,

Mago Torres/Google News Initiative (00:39:42):

You can search of course you can use the search space if you want to look for something very specific, if somebody gives you or you obtain a set of documents from a FOIA request and you want to put them there, you can start let’s say public meetings in a specific county. I used to live in Chicago for five years also in Cook County in Illinois and I have all of them and I want to see how many of them mentioned education. I can do the search education but maybe I want to see those who talk about education but doesn’t touch Chicago public schools. So I can, I would do the search advanced search and I can say but not Chicago public school. So just when you are doing the searches, just how you can use the same operators that you will do in a regular search to narrow what you’re looking for in understanding your documents.

(00:40:54):

There’s something really cool and one of the things I really like to do is to look into companies, public companies, annual reports. I think there’s a gold mine of information there and understanding for example, how a company will express, if I’m looking into oil companies and how one company, I think it’s BP in the uk may talk about their risk in their annual reports versus chairman annual reports. It’s just like using different words. It’s just like how can I understand? So if you know some keywords, a search word pinpoint is going to recognize similar matches. So lemme show you this one. So I know this is a NASA collection and I know it talks about something related with the moon. So I’m going to search for moon here and what pinpoint is going to recognize actually the word lunar. So if you have other terms depending on what is in your collection and you do search right terms and pinpoint recognize there similar matches. And I think when we are doing some reporting and trying to figure out some of the words and that the documents contained is very helpful to do these kind of searches. So I think if this is something very helpful, I’m looking here because I want to be sure I have all the features. Another one, how are we going? Going well

(00:42:32):

Perfect. If you have questions, please let me know. I

Lia DeGroot | CQ Roll Call (00:42:39):

Question. Question,

Mago Torres/Google News Initiative (00:42:41):

Yeah. Yes, yes, yes.

Lia DeGroot | CQ Roll Call (00:42:43):

Hi, I’m Lia. I’m with CQ roll call and a lot of what I do is compare two different versions of the same bill. Is there a PDF compare tool where it shows you what’s different in each document?

Mago Torres/Google News Initiative (00:43:00):

Yes, I have the answer for you. That is in the AI features. So let me finish with the non-AI features and we’ll go there. Sounds good. Fantastic. Okay. Something that I mentioned is also the handwriting recognition. Something I didn’t mention. If you have a small collections, these one has images and different elements. You can just click here and you’re going to see the next document. I want you to see here that it says pro three, this is a PDF, this is handwritten and I can already select it. That’s how handwritten recognition happens. Sometimes depending on the handwritten is more complex, it could be more difficult. But same again, if you have a unique case and you want to figure it out, you can reach out to the team because the more information they have is how also the tool can improve and support your work. So here for example, what I’m going to do with my keyboards, you’re going to say copy, I just like command copy and I have this Google doc, I’ll share it in a minute. I just want to be sure that my computer is not going to crash here I have this Google doc and I want you to see I did copy and I am going to put paste and it recognize the text. So this is just for you to see how pinpoint can recognize handwritten in documents. And then let’s see the audio feature. This is the file. So the way it’s going to make pause is going to recognize audio is like when there are pauses in the person that is speaking. So here I’m just going to put it all and it said like

Speaker 11 (00:45:26):

This is Neil Armstrong, my fellow Apollo 11 crewman and I would like to invite you to take a close look at our spacecraft Columbia while it is in your state Columbia.

Mago Torres/Google News Initiative (00:45:37):

So that’s the text. Let’s say that he didn’t say this is Neil Armstrong. What he said was these are Neil Armstrong and Malo address and I can save it and this is going to appear, it’s not going to change the original audio because that’s not what it says. But sometimes when you can identify a mistake you can fix it. And something that I work a lot with entity extraction and something that happens here is because I added the name and let’s say that I corrected the document and I look for Mago Torres appears as one of the entity, one of the entities here. So if there’s something that could be not, there’s a problem with the transcription and you fix it, it’s also going to be recognized across the documents and when you do your searches. But we know that, oops, I’m going to fix that because I was definitely not there. So that’s how you work with the edit. Can edit the and fix the transcription. There’s something else. I’m going to do it here. It happens equally in text and transcription and transcribe files. I select this sentence.

(00:47:27):

I want you to remember every we see this little star is an AI feature. I’m going to leave it this for later. We just finished this and we will go to the AI features. But this is something that I found extremely helpful. That is the highlight and get the link. Let’s say that this is my draft. Let’s say that I’m working on this draft that is a very complex draft that I use that a specific quote and I’m going to insert it here as a footnote. And what I can do, I want these very specific quote and very specific document for the fact checker to see it. And if you see here, it generates the very specific link to very specific sentence in the document that I share. So if I go to the document, it is going to take the person that link’s going to take it specifically to that very specific quote.

(00:48:39):

When we are working with documents that have 300 pages, it’s just like a lifesaver, it saves time and we do have, depending on the type of a story, but it’s very complex and it is not only the editor, the ones that have to see it but also the fact checker and therefore go to the lawyer review whatever all the evidence is containing one place. And I think that when you work in collaborations, when you have a team of people working on something, do you want to be very accurate is having this alternatives just like I just found it extremely helpful.

(00:49:20):

Okay, let’s go to the AI features. We were here. Oh and when you search by date you can put the date when you search by date. If you say well I want to look into a very specific timeframe within these documents, pinpoint is going to recognize even though the search indicates to use a year, month, day format in the document is going to be able to recognize if they’re in different format within the documents and it’s going to be very helpful. So now let’s go to the AI features. And this is the question I’m actually going to change collection because I think it’s just understanding numerous when I was mentioning for example, I like to look into annual reports. I think there’s a lot of information. So I this is a collection that has four annual reports for Chevron.

(00:50:24):

These are the glossy reports and this is how it looks like, right? There’s a lot of information there and it’s already telling me that this one has 104 pages. If I go to the other one it’s 116 pages, same for 2022 and 114 pages for 2023. But what if I want to, the AI features can tell me I can compare. So I want to look into the documents between the annual reports between 2021 and 2023 and I want to compare how I can give it a topic or not or I can give it and that’s going to be the key. And this is to the question that was made, I’m sorry, I’m sorry I missed the name but this is how you can select two documents to compare. In this case these are not the same, it’s two different years but it’s like these annual reports follow the same format.

(00:51:32):

So I want to know how I remember what risks and I want to understand what are the risks where companies can talk about environmental risks like these type of companies where legal risks, risk, financial risk, whatever. So I’m asking and I’m just telling that I want to understand what risks are described by year and I just highlighted two documents here. I’m very specific and it’s going to tell me here pinpoint is going to give you different reminders of what you’re doing here. I can say it is telling me you are asking to compare between two documents and if I remove the check marks is going to tell me in the entire collection or actually I’m going to ask only for three years. So you can mark it like that and I’m going to say compare.

(00:52:45):

It’s going to take a little bit of time. It’s telling me that I can keep going, they will give me my answers in a little bit. Actually I already had a comparison before and I said, and I did a comparison between two documents. I asked for environmental impact and this is also I think something very important that we care about when we’re doing journalism is that we want to know exactly where the information comes from. So I asked to compare two documents here and with the topic of environmental impact and this is the response that generative AI that Germany is giving to me. It’s just like this is both documents state this and this and this, but it’s giving me the source of information. So what this number is telling me is this sentence, we are getting it from two places and if I click there it’s telling me from where it is coming.

(00:53:48):

So it’s always going to indicate why it is saying. So I think even when we want to, we’re using the generative AI features but we want to go, it’s going to always link us to the original source and it’s always within our collection of documents. So this is one way in which you can use Gemini to compare to documents and what you can do if you have two documents and depending on the comparison better with the type of context and I will be very interested if you use it and how is being helpful? If you want to share with me that will be amazing.

(00:54:29):

Oh my god, my idea that you work on your prompt, just like if you can make it more specific, I want you to compare this two documents organizes like these that you test that type of information and this is comparison. If you only have an answer, if you are only looking for answers, you can also type your question and ask it to answer. We did, we did manual labeling of collections with the NASA demo and Kevin asked, I did these labels out manually because there was something very specific that I knew that I was there. But you can also try the generative features to label documents. My condition is that the documents that contain

Speaker 3 (00:55:39):

Photograph

Mago Torres/Google News Initiative (00:55:47):

I think I already created, yeah, photo two I’m going to put it and I’m going to say identify, but this case I’m using Gemini documents that contain a photograph because I already know that there are some here and I want to say label, the generative AI feature is working

Speaker 3 (00:56:25):

Interesting.

Mago Torres/Google News Initiative (00:56:32):

I know there are labels so

(00:57:03):

Now it labeled, it identified. So I changed my prompt and identify 20 documents that have photos. I didn’t have to do it manually. What we always say when you are using generative features, nothing goes unchecked. So be sure that when you have filters just like, oh is this accurate? This is pretty accurate in some cases because of this is an old document, it was an A photo and I can decide, well this is going to be helpful to keep us an image or not. But this is a very long document but contains images. So that’s another way in which you can use the label documents. If you have a specific type of documents and you want to have a timeline, you can also create it. I was showing earlier maybe I want to see what a company is saying of the timeline of its growth. So I can be very specific in market or I can say I don’t want want a timeline, I don’t want to focus on everything and anything. Just create a timeline and get it. The more your documents, the more specific and you will be able to decide how to use these features. You can ask for summary and also specific questions to your collections when you ask. Yeah, no it’s not here.

(00:58:50):

So these are some of the how you can start seeing and I want to be very mindful of the time and we have three minutes left. There’s one thing that I think is very important. I have worked always in collaboration and one thing I was mentioning when you can share your document with others, so let’s say that I am working with a reporter here or with an organization, you can share your collection and when you share your collection, I’m going to put my friend Erica, let’s say that I want to share with her. I can decide what type to people. If you want to share like oh I want to share it with my collaborator. We are writing the story together, we are adding documents to the collection. We are highlighting, selecting, asking questions. We are highlighting things, elements are going to the story. You add it as an editor, you want someone that is only going to be reviewing, you don’t want them to change anything, you use the viewer.

(01:00:11):

So that’s very important when you are sharing documents and collaborating with others. When you publish, when you want to make a collection public, and I think this is really beautiful is that you can change the name, you can add a description and if you are already write stories that you use that collection, you can add the title and the link here. So whoever has access to your once this repository of public collections that we were seeing, the one with the AP and newsrooms around the world, having uploading documents, there is specific stories you can start adding them so whoever goes to the collection is going to be able what stories have been done with this. And I think it’s just another way of giving that space for your stories to be seen. One final thing before I forget, and this is because I mentioned it, there is something that I said when we highlight it is right there, but I’m going to use a different highlight.

(01:01:24):

You can select when you are reading a document, let’s say you decide to upload a transcription of a public speaking, how is it called? The state of the Union, right? You upload the transcription and you want to understand you have one sentence from the state of the union, I have this annual report. I select the sentence and I use the AI feature to explain in context. And what it’s going to do is going to create an explanation of that very specific sentence according to the context of the document. What I am asking, again when we have very, very large documents and we are just starting to make sense of something we want to, it’s very helpful to start approaching to understanding and here look all the ways in which it’s giving me 15 sources within the same document to understand that very specific sentence that happens also to be the slogan of a company. So it’s just a very interesting way of starting exploring documents that otherwise will take us a lot of time. I want to stop here because of the time I haven’t had a chance to show you getting structured information. It will take me 10 minutes or less, but I just want to be sure if you want to see it or if you have a question or a reaction or what do you think of what Pinpoint offers you so far?

Kevin Johnson/NPF (01:03:07):

Anybody have any questions?

Sydney Clark/NPF (01:03:10):

I have one. This is Sydney with National Press Foundation. Are all of the elements and features of Pinpoint Free or are there some parts that you have to pay for?

Mago Torres/Google News Initiative (01:03:21):

Everything is

Sydney Clark/NPF (01:03:22):

Free. Awesome, thank you.

Kevin Johnson/NPF (01:03:25):

And this is Kevin. In terms of help when it’s needed, how do you go about doing that and how accessible are you all from day to day?

Mago Torres/Google News Initiative (01:03:39):

Yeah. You mean help from the pinpoint team?

Speaker 3 (01:03:45):

Yes.

Mago Torres/Google News Initiative (01:03:45):

Yeah, there’s an email. You can send an email and in terms of me as a trainer, I’m going to include a link to my calendar and if there’s something you want to chat about the specific, I’ll be happy to schedule a 30 minute call. And if we need more time, that’s what we offer. We trainers are our work is to help and teach you the tools and the pinpoint team is very super responsive via email.

Speaker 3 (01:04:24):

Yeah.

Mago Torres/Google News Initiative (01:04:25):

Do you think, I know that only one person knew the tool, so I am curious to know what are your first reactions since you are seeing it and learning a little bit more for the first time. Would you want to give it a shot? Try have.

Kevin Johnson/NPF (01:04:49):

Yeah, a lot of nods. I think it’s one of these things that you don’t know how effective it could be until you get your hands on it, right? So I think that’s probably where a lot of folks stand.

Mago Torres/Google News Initiative (01:05:05):

Yeah, yeah, I understand. Yeah, absolutely. And there are a few more tools that can be helpful for you. One, we have a training on Google Trends if you want to, how use it for understanding what people are searching and also we have trends on Gemini and Notebook Lamb that this are another generative ai. Well, Gemini is the Google’s chatbot and notebook lamb. There is an amazing tool also for analyzing documents from different features than Pinpoint. And also the three of them share some overlaps. I think it’s very interesting to understanding what serves to our work and when we see all these features. If you are going to test it and use it for the first time, I always say just to start one by one, don’t try to do everything at the same time. Test it and understanding what will work better for you depending on what you need.

Kevin Johnson/NPF (01:06:09):

Okay, well great. Well we can’t thank you enough, Margo. Thank you so much for this long distance instruction. Yeah, of course. All is well in California and thank you. Really

Mago Torres/Google News Initiative (01:06:25):

Appreciate it. Yeah, absolutely. I’ll send the email Kevin, and just for all the information to be shared.

Kevin Johnson/NPF (01:06:31):

Okay, wonderful. Thank you.

Mago Torres/Google News Initiative (01:06:34):

Thank you.