Key Moments
The New "Bill Of Rights of Information Society"
Google TalksKey Moments
The 'Bill of Rights of the Information Society' aims to deliver the right info to the right people at the right time, but current tech falls short, especially for illiterate populations who need problem-solving answers, not just links.
Key Insights
The 'Bill of Rights of the Information Society' includes six 'rights': right information, right people, right time, right language, right medium, and right level of detail.
Current search engines often fail to eliminate massive redundancy, providing numerous links to similar information rather than distinct perspectives or consolidated answers.
The concept of 'right time' information is not just about immediate delivery but also about 'just-in-time' delivery, avoiding information overload and forgetting.
Machine translation tools are still not 'good enough,' and even distinguishing between languages based on minimal text (like a book title) presents a challenge.
Information needs vary greatly by user context; for example, a child's query about heart health requires different results than a doctor's, highlighting the 'right people' challenge.
Summarization needs to go beyond single-paragraph abstracts to offer hierarchical levels of detail, from a headline to an executive summary, catering to different information needs and time constraints.
Defining the 'Bill of Rights of the Information Society'
The core mission for organizing the world's information, as articulated by Jaime Carbonell and expanded upon by Raj Reddy, is to deliver the 'right information to the right people at the right time, in the right language, and at the right level of detail.' This framework, jokingly termed the 'Bill of Rights of the Information Society,' highlights the immense research challenges that remain despite significant technological advancements. An additional crucial 'right' is the 'right medium,' which is particularly vital for illiterate populations who may require audio and video formats over text. This comprehensive set of 'rights' implies a broad research agenda encompassing search engines, information classification, anticipatory analysis, machine translation, and speech/video input/output.
The challenge of 'right information' and combating redundancy
Defining 'right information' is complex, as it is highly contextual. For a student, it might be class material; for someone getting married, it's planning resources. The current web publishing paradigm treats all information sources equally, making it difficult to discern trusted content from marketing or misinformation. Search engines often suffer from massive redundancy, returning numerous similar links (e.g., multiple Yahoo email promotions when searching for 'web-based email'). The ideal scenario would offer a comparison of distinct services like Yahoo, Gmail, and MSN, rather than overwhelming users with repetitive results. Research is needed to move beyond mere relevance and popularity to identify and present maximally diverse and novel information.
Novelty detection and tracking information evolution
Identifying novel information is crucial. This involves detecting when new terms or events emerge, like the word 'Katrina' suddenly becoming a prominent news topic. Techniques like first story detection (FSD) help identify these emerging events. Once detected, tracking the evolution of a story over time is also vital. For instance, casualty figures for disasters like Katrina or 9/11 change daily. Simply presenting the latest number can be misleading. The system needs to handle non-monotonic reasoning—where facts change over time—to eventually stabilize on a reliable piece of information or acknowledge when a fact is no longer relevant. This also extends to understanding the context and evolution of named entities, such as individuals, locations, and organizations mentioned in texts.
Named entity extraction for deeper understanding
Named entity extraction (NEE) is essential for understanding who is mentioned, where, and what companies or products are involved. This is particularly useful when encountering unfamiliar names or entities, such as obscure prizes like the Yukawa or Honda prizes, which might be important in one region but unknown elsewhere. The system needs to identify entities and their roles (e.g., who participated in a meeting, who was the host). While knowledge bases can help categorize known entities, identifying new or foreign names and inferring relationships requires sophisticated techniques like finite state transducers and statistical learning. This process transforms unstructured text into semi-structured data, akin to populating a database, which is critical for answering specific queries about relationships ('who does what to whom').
The complexity of identifying the 'right people'
Identifying the 'right people' to receive information is as challenging as defining the 'right information.' A system needs to understand the user's context—their age, expertise, or affiliation (e.g., a child seeking health information versus a doctor). Simply providing raw search results is insufficient; the system must infer the user's informational needs. While humans can easily distinguish a doctor from a child, machines struggle with this. Categorizing and classifying text to infer suitability for different user groups is a significant research area, moving beyond manual categorization or simple machine learning to more sophisticated hierarchical classification.
Optimizing information delivery through the 'right time'
The 'right time' for information delivery is not necessarily 'as soon as possible.' Sending information too early can lead to it being forgotten or lost, a problem often seen in traditional education where subjects are taught long before they are needed. The ideal is 'just-in-time' information, provided precisely when it is required. This includes proactive or anticipatory computing, where potential results are pre-computed and stored, ready for immediate retrieval. While alerts and reminders (like breaking news) can be useful, they can also be intrusive and harmful if delivered at inconvenient moments, underscoring the need for user control over information timing.
Overcoming barriers in 'right language' and 'right medium'
Accurate machine translation remains a difficult problem, with current tools often falling short. Even basic tasks like language identification from minimal text (e.g., a book title) can be challenging, especially for less common languages. For illiterate populations, the 'right medium' is critical; information must be accessible via audio and video, not just text. This requires advancements in speech and video processing. Transliteration, converting sounds across different scripts while preserving meaning, is also a significant challenge, particularly for languages with diverse writing systems. Hybrid or multi-engine translation systems are being explored to combine the strengths of various approaches.
Achieving the 'right level of detail' through advanced summarization
Providing information at the 'right level of detail' necessitates sophisticated summarization technologies that go beyond simple abstracts. Users often need tiered summaries: a headline, a short abstract, an executive summary, and potentially longer summaries, allowing them to gauge relevance and depth without reading entire documents. For highly focused queries, especially from users who may be illiterate, the need is not just for summarized information but for the computer to actively solve a problem and provide a direct answer, akin to active knowledge rather than passive information retrieval. This could involve systems that understand a user's health issue and recommend treatment, rather than just providing links to medical articles.
Mentioned in This Episode
●Software & Apps
●Companies
●Organizations
●Concepts
●People Referenced
Common Questions
The core principles, inspired by Haime Carbonell, focus on providing the right information to the right people, at the right time, in the right language, and at the right level of detail, with the added consideration of the right medium (audio/video for non-literate individuals).
Topics
Mentioned in this video
Colleague of Raj Reddy and head of the Language Technology Institute at CMU, who inspired the talk's theme by articulating the 'Bill of Rights of the Information Society'.
One of the inventors of the IP protocol and the internet, with whom the speaker discussed the concept of an 'AAA web' (authentic, archival, and always available information).
Distinguished computer scientist, recipient of the Turing Award and other accolades, who is presented as the 'strong force' of the computer science community and a proponent of bringing computing to the developing world. He founded the PCTV project and the Million Books project.
A prize awarded in Japan that Raj Reddy received, highlighting the need for named entity extraction to identify and understand the significance of such awards, especially to those unfamiliar with them.
A physics concept mentioned in the introduction regarding the force that binds quarks together, used metaphorically to describe Raj Reddy's role in the computer science community.
An award received by Raj Reddy.
A set of principles articulated by Haime Carbonell regarding the provision of information, encompassing the right information to the right people at the right time, in the right language, and at the right level of detail, with an additional focus on the right medium.
The most important prize in computer science mentioned as having been won by Raj Reddy.
Mentioned as a protocol co-invented by Bob Kahn, in the context of discussing information authenticity and archival.
Discussed as a crucial research area for providing information in the right language, with current tools being imperfect but improving.
A prize won by Raj Reddy.
Mentioned as a system co-invented by Bob Kahn, related to the discussion on information authenticity and archival (AAA web).
A character encoding standard that, according to the speaker, was designed incorrectly for Indian languages regarding sound versus letter representation.
A project founded by Raj Reddy.
A project founded by Raj Reddy, aiming to digitize a million books, which involved securing agreement from the presidents of India and China.
Cited as an example of trusted sources of information that, once published, cannot be changed, unlike web sources.
An institute at CMU, led by Haime Carbonell, focused on language processing, which inspired the talk's central theme.
The speaker corrects himself, stating that the Nobel Prize winners Pitzer and Wilczek were from Caltech, not Wilck as initially misspoken.
Mentioned as a funder for research on 'orphan languages' and obscure languages that commercial organizations like Google might not prioritize.
Mentioned in the context of diverse reporting of news stories, contrasting how events might be presented differently in Palestinian newspapers compared to others.
Mentioned as a provider of web-based email and as an example of a company that previously used manual category assignment for its services. Also mentioned in relation to search results for web-based email.
The company where the talk is being given, with the speaker's mission being to encourage Google to do more in organizing world information. The company's search engine and its beginnings are also mentioned.
Mentioned as a provider of web-based email that should be included in comparative search results.
Mentioned as a company whose reporting might differ from Google's, highlighting the concept of varied perspectives on the same story.
Mentioned as a past effort that might be related to solving problems rather than just providing information, though it apparently did not gain significant traction.
Mentioned as a web-based email service that should be included in search results, alongside Yahoo and MSN.
Mentioned as a predecessor to current book digitization efforts, with the Million Books Digital Library project being an earlier initiative.
Mentioned in the context of diverse reporting of news stories, contrasting how events in Israel might be presented differently in Palestinian newspapers.
Mentioned in the context of disaster response and tracking data, where the number of casualties can change over time, illustrating the challenge of non-monotonic reasoning in information retrieval.
More from GoogleTalksArchive
View all 92 summaries
58 minEverything is Miscellaneous
54 minStatistical Aspects of Data Mining (Stats 202) Day 7
45 minKey Phrase Indexing With Controlled Vocabularies
63 minMysteries of the Human Genome
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free