3 ms·

Ehm... The semantic web as an idea was/is a totally different thing: the idea is the old libraries of Babel/Bibliotheca Universalis by Conrad Gessner (~1545) [1

by kkfx 2y ago

Ehm... The semantic web as an idea was/is a totally different thing: the idea is the old libraries of Babel/Bibliotheca Universalis by Conrad Gessner (~1545) [1] or the ability to "narrow"|"select"|"find" just "the small bit of information I want". Observing that a book it's excellent to develop and share a specific topic, it have some indexes to help directly find specific information but that's not enough, a library of books can't be traversed quick enough to find a very specific bit of information like when John Smith was born and where. The semantic web original idea was the interconnection of every bit of information in a format a machine can travel for a human, so the human can find any specific bit ever written with little to no effort without having to humanly scan pages of moderately related stuff. We never achieve such goal. Some have tried to be more on the machine side, like WikiData, some have pushed to the extreme the library science SGML idea of universal classification not ended to JSON but all are failures because they are not universal nor easy to "select and assemble specific bit of information" on human queries. LLMs are a, failed, tentative of achieve such result from another way, their hallucinations and slow formation of a model prove their substantial failure, they SEEMS to succeed for a distracted eye perceiving just the wow effect, but they practically fails. Aside the issue with ALL test done on the metadata side of the spectrum so far is simple: in theory we can all be good citizens and carefully label anything, even classify following Dublin Core at al any single page, in practice very few do so, all the rest do not care, or ignoring the classification at all or badly implemented it, and as a result is like an archive with some missing documents, you'll always have holes in information breaking the credibility/practical usefulness of the tool. Essentially that's why we keep using search engines every day, with classic keyword based matches and some extras around. Words are the common denominator for textual information and the larger slice of our information is textual. [1] https://en.wikipedia.org/wiki/Bibliotheca_universalis https://en.wikipedia.org/wiki/Bibliotheca_universalis

DrScientist 2y ago
The problem I find with semantic search is first I have to read and understand somebody elses definitions before I can search within the confines of the ontology. The problem I have with ML guided search is the ML takes web average view of what I mean, which sometimes I need to understand and then try and work around if that's wrong. It can become impossible to find stuff off the beaten track. The nice thing about keyword and exact text searching with fast iteration is it's my mental model that is driving the results. However if it's an area I don't know much about there is a chicken and egg problem of knowing which words to use.
- kkfx 2y ago
  Personally I think the limitation of keyword search it's not in the model per se but in the human langue: we have synonymous witch are relatively easy to handle but we also have gazillion of different way to express the very same concept that simply can't be squeezed in some "nearby keyword list". Personally I notes news, importing articles in org-mode, so I have a "trail" of the news I think are relevant in a timeline, sometimes I remember I've noted something but I can't find it immediately in my own notes with local full-text search on a very little base compared to the entire web, simply because a title does express something with very different words than another and at the moment of a search I do not think about such possible expression. For casual searches we do not notice, but for some specific searches emerge very clear as a big limitation, however so far LLMs does not solve it, they are even LESS able to extract relevant information, and "semantic" classifications does not seems to be effective either, a thing even easier to spot if you use Zotero and tags and really try to use tags to look for something, in the end you'll resort on mere keyword search for anything. That's why IMVHO it's an unsolved so far problem.
  - DrScientist 2y ago
    For me the search problem isn't so much about making sure I get back all potentially relevant hits ( more than I could ever read ) , it's how I get the specific ones I want... So effective search more about excluding than including. Exact phrases or particular keywords are great tools here. Note there is also a difference between finding an answer to a particular question and finding web pages around a particular topic. Perhaps LLM's are more useful for the former - where there is a need to both map the question to an embedding, and summarize the answer - but for the latter I'm not interested in a summary/quick answer, I'm interested in the source material. Sometimes you can combine the two - LLM's for a quick route into the common jargon, which can then be used as keywords.
    - kkfx 2y ago
      > there is also a difference between finding an answer to a particular question and finding web pages around a particular topic. That's exactly the Gessner's problem and unimplementable so far solution: we have information packed in various form, when we want specific bits it's hard do narrow enough available information packages (books, articles, post etc). For common stuff in the vast see of the web we tend to end up very frequently on exactly or nearly exactly what we want, but for less common stuff there is a big problem. Just knowing the historical temperature of a small place where many have recorded temperatures, but still no one have create a table and a graph or daily minima and maxima for let's say the 10 years period we want it's hard to find an answer. For some kind of data we have wikidata, for some others we have public datasets from various sources, but still no easy way to locate and narrow them. LLM are in theory an answer, unfortunately not in practice first for freshness problem (you have to train a model, ingesting new information is still training, far longer than a search engine crawler) and training itself witch is like the opposite of semantic web: instead of having the author prepare the information for machines we have third party who do it en masse, unfortunately they tend to be even less precise than semantic web mean authors PLUS models hallucinations and essentially no ability to tell where their answer came from. That's why we still miss semantic search...

by kkfx 2y ago

DrScientist 2y ago

The problem I find with semantic search is first I have to read and understand somebody elses definitions before I can search within the confines of the ontology. The problem I have with ML guided search is the ML takes web average view of what I mean, which sometimes I need to understand and then try and work around if that's wrong. It can become impossible to find stuff off the beaten track. The nice thing about keyword and exact text searching with fast iteration is it's my mental model that is driving the results. However if it's an area I don't know much about there is a chicken and egg problem of knowing which words to use.

kkfx 2y ago
Personally I think the limitation of keyword search it's not in the model per se but in the human langue: we have synonymous witch are relatively easy to handle but we also have gazillion of different way to express the very same concept that simply can't be squeezed in some "nearby keyword list". Personally I notes news, importing articles in org-mode, so I have a "trail" of the news I think are relevant in a timeline, sometimes I remember I've noted something but I can't find it immediately in my own notes with local full-text search on a very little base compared to the entire web, simply because a title does express something with very different words than another and at the moment of a search I do not think about such possible expression. For casual searches we do not notice, but for some specific searches emerge very clear as a big limitation, however so far LLMs does not solve it, they are even LESS able to extract relevant information, and "semantic" classifications does not seems to be effective either, a thing even easier to spot if you use Zotero and tags and really try to use tags to look for something, in the end you'll resort on mere keyword search for anything. That's why IMVHO it's an unsolved so far problem.
- DrScientist 2y ago
  For me the search problem isn't so much about making sure I get back all potentially relevant hits ( more than I could ever read ) , it's how I get the specific ones I want... So effective search more about excluding than including. Exact phrases or particular keywords are great tools here. Note there is also a difference between finding an answer to a particular question and finding web pages around a particular topic. Perhaps LLM's are more useful for the former - where there is a need to both map the question to an embedding, and summarize the answer - but for the latter I'm not interested in a summary/quick answer, I'm interested in the source material. Sometimes you can combine the two - LLM's for a quick route into the common jargon, which can then be used as keywords.
  - kkfx 2y ago
    > there is also a difference between finding an answer to a particular question and finding web pages around a particular topic. That's exactly the Gessner's problem and unimplementable so far solution: we have information packed in various form, when we want specific bits it's hard do narrow enough available information packages (books, articles, post etc). For common stuff in the vast see of the web we tend to end up very frequently on exactly or nearly exactly what we want, but for less common stuff there is a big problem. Just knowing the historical temperature of a small place where many have recorded temperatures, but still no one have create a table and a graph or daily minima and maxima for let's say the 10 years period we want it's hard to find an answer. For some kind of data we have wikidata, for some others we have public datasets from various sources, but still no easy way to locate and narrow them. LLM are in theory an answer, unfortunately not in practice first for freshness problem (you have to train a model, ingesting new information is still training, far longer than a search engine crawler) and training itself witch is like the opposite of semantic web: instead of having the author prepare the information for machines we have third party who do it en masse, unfortunately they tend to be even less precise than semantic web mean authors PLUS models hallucinations and essentially no ability to tell where their answer came from. That's why we still miss semantic search...