4 ms·

Data. Currently, the web is still a big collection of text documents. What is not clear to us right now because we're knee-deep in it is is how deeply sub-opti

Data. Currently, the web is still a big collection of text documents. What is not clear to us right now because we're knee-deep in it is is how deeply sub-optimal that is. The number of people going to the Google homepage is declining because people are now just going directly to the database that represents what they're looking for. Looking to buy a home? You go to Zillow or Trulia and query their database. People want data, not text. Text is for novels and poetry, not for things you want to find. 95%, almost all, of the web was posted to be found -- it wants to be data. What will come next is the web as a database and Google will no longer be a player there, new companies will be.

BjoernKW 10y ago
The problem with that is that natural language - written or spoken - is the natural user interface for humans. To a certain extent we can learn how to input data in a structured manner a machine can easily process: - the limited subset of human language that makes sense as search engine input - GUIs - pretty much every input form in client software However, for anything above a certain level of complexity this requires domain knowledge and often even for users with experience in that domain it's still an error-prone process. The web as a general-purpose database is a vision that has yet to materialize. It's not that hasn't been tried before, both in a top-down (Semantic Web) and bottom-up (microformats, Wikis) fashion. It's not that those attempts have been complete failures: The tools they brought about - particularly Wikis - are a vital and highly useful part of the web. Still, they're mostly limited to niche or domain-specific data and we still need search engines and natural language understanding to make sense of the largely unstructured free-form text that's stored in Wikis. That's why I still see the need for a player such as Google that in one way or another makes human language comprehensible to machines. This might result in something like the web-as-a-database but it isn't all that different from what the likes of Google and Facebook are already doing today, is it?
- david927 10y ago
  As for natural language interfaces, you can, right now, go to Google and type in "two-bedroom homes in Paradise, California". It won't try to answer it. It will instead send you to a site that is a database, right? And as I mentioned, numbers from Google's homepage are declining specifically because people are skipping that step. The web as a general-purpose database is a vision that has yet to materialize. Yes, sure. I guess that's my point: its materialization will be the basis of the next cycle. I have concrete grounds to believe that it's on its way and it will be tremendously exciting because it's the beginning of the final form. The data will get more complex, more contextual, and more refined, but it will always be data. The astonishing thing about new paradigms is that it's like suddenly being able to see; entire new dimensions open up; things that we didn't even consider before become suddenly cake.
  - BjoernKW 10y ago
    > numbers from Google's homepage are declining specifically because people are skipping that step. That's only if people know already which database to query. Moreover, to some extent, Google's already trying to provide a direct answer to user queries (flights, directions, queries for which a Wikipeda / Wiktionary abstract is the best hit). Doing a "SELECT * FROM restaurants ORDER BY review DESC" is simple, discovering the database to execute that query against is still difficult. We'd need some sort of service discovery to automate that beyond making machines understand natural language.

BjoernKW 10y ago

The problem with that is that natural language - written or spoken - is the natural user interface for humans. To a certain extent we can learn how to input data in a structured manner a machine can easily process: - the limited subset of human language that makes sense as search engine input - GUIs - pretty much every input form in client software However, for anything above a certain level of complexity this requires domain knowledge and often even for users with experience in that domain it's still an error-prone process. The web as a general-purpose database is a vision that has yet to materialize. It's not that hasn't been tried before, both in a top-down (Semantic Web) and bottom-up (microformats, Wikis) fashion. It's not that those attempts have been complete failures: The tools they brought about - particularly Wikis - are a vital and highly useful part of the web. Still, they're mostly limited to niche or domain-specific data and we still need search engines and natural language understanding to make sense of the largely unstructured free-form text that's stored in Wikis. That's why I still see the need for a player such as Google that in one way or another makes human language comprehensible to machines. This might result in something like the web-as-a-database but it isn't all that different from what the likes of Google and Facebook are already doing today, is it?

david927 10y ago
As for natural language interfaces, you can, right now, go to Google and type in "two-bedroom homes in Paradise, California". It won't try to answer it. It will instead send you to a site that is a database, right? And as I mentioned, numbers from Google's homepage are declining specifically because people are skipping that step. The web as a general-purpose database is a vision that has yet to materialize. Yes, sure. I guess that's my point: its materialization will be the basis of the next cycle. I have concrete grounds to believe that it's on its way and it will be tremendously exciting because it's the beginning of the final form. The data will get more complex, more contextual, and more refined, but it will always be data. The astonishing thing about new paradigms is that it's like suddenly being able to see; entire new dimensions open up; things that we didn't even consider before become suddenly cake.
- BjoernKW 10y ago
  > numbers from Google's homepage are declining specifically because people are skipping that step. That's only if people know already which database to query. Moreover, to some extent, Google's already trying to provide a direct answer to user queries (flights, directions, queries for which a Wikipeda / Wiktionary abstract is the best hit). Doing a "SELECT * FROM restaurants ORDER BY review DESC" is simple, discovering the database to execute that query against is still difficult. We'd need some sort of service discovery to automate that beyond making machines understand natural language.