TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Show HN: Shift-Ctrl-F: Search a webpage with natural language and TensorFlowJS

143 pointsby yoavzover 4 years ago

16 comments

kevincoxover 4 years ago
I think more natural language and context is the next huge step in search. The README has a good example where you find a section comparing two things. An example I run into often is trying to find emails or texts about an event. I know the date that the event occurred but I might have said "tomorrow", "tuesday", "the 25th", "2020-08-25" or "yesterday". These all refer to the same date can could be indexed, however Now I need to search for all of these with different date restrictions to find the hits and not show the misses.
评论 #24275985 未加载
Der_Einzigeover 4 years ago
I&#x27;ve been waiting for someone to do a proper semantic search plugin in a browser for awhile. There was one awhile back called... Fuzbal ... which used word2vec and was good but has not been updated. You&#x27;ve implemented a more question-answer based approach. This is awesome!<p>I think that the real innovation will be when users are given exposure to lots of different models, and have the pros and cons of these models are properly explained to them. Maybe I want to use this on specialized bio-medical literature and would be better off with a model fine-tuned in that domain instead of on Squad.<p>Also, shameless self-plug, I wrote a system that does extractive summarization&#x2F;highlighting of documents which is in principle very similar to what is going on here (<a href="https:&#x2F;&#x2F;github.com&#x2F;Hellisotherpeople&#x2F;CX_DB8" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;Hellisotherpeople&#x2F;CX_DB8</a>). For awhile, I had a hosted, web accessible version of this system available to make it easy to show it off to interviewers. It could highlight the important parts of a web-page based on a user query at either the word, sentence, n-gram, or paragraph level. I figured that the next step was to make it a browser extension. I simply wasn&#x27;t proficient enough in JS and at the time I was working on this, quantized&#x2F;pruned models were slightly less good. I firmly believe that making high quality semantic search work everywhere will be an extreme (and obvious) step-forward for most peoples daily tasks. What a brave new world we are entering!
评论 #24275971 未加载
lbjover 4 years ago
Wow. Now that’s an innovative and brilliant way to improve one our oldest tools. Certainly could by relevant in a general sense for much more than browsing
评论 #24275062 未加载
adrianmonkover 4 years ago
Interesting idea for sure. I wasn&#x27;t able to understand much from the demo image, though. The animation is fast, and all I can see about the result is that the word &quot;lower&quot; is highlighted&#x2F;matched. I was hoping to get an idea of what results it finds and how relevant they are to the search.
评论 #24273524 未加载
ReD_CoDEover 4 years ago
Wow, great!<p>I&#x27;m looking for an open source solution to find algorithm names inside the academic articles (normally PDF), and perhaps on the web too<p>Is there any suggestion?
评论 #24277627 未加载
paraschopraover 4 years ago
Isn’t this exactly like what Google released as open source a couple of months ago <a href="https:&#x2F;&#x2F;github.com&#x2F;tensorflow&#x2F;tfjs-models&#x2F;tree&#x2F;master&#x2F;qna" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;tensorflow&#x2F;tfjs-models&#x2F;tree&#x2F;master&#x2F;qna</a>
评论 #24279011 未加载
roland-sover 4 years ago
OpenAI API has a similar demo, the Wikipedia one at <a href="https:&#x2F;&#x2F;openai.com&#x2F;blog&#x2F;openai-api&#x2F;" rel="nofollow">https:&#x2F;&#x2F;openai.com&#x2F;blog&#x2F;openai-api&#x2F;</a>
de6u99erover 4 years ago
Does the use TensoflowJS mean that search is being performed locally?
评论 #24275084 未加载
dvaunover 4 years ago
This is definitely an interesting project. I&#x27;ll give it a shot with Chrome the next time I&#x27;m scouring Reddit or HN for information when I&#x27;m doing research for a project
dzhiurgisover 4 years ago
Could be an awesome plug-in for IntelliJ (which is already super awesome compared to macos spell checker and text navigation)
krickover 4 years ago
Did someone try it? Is it actually good?
评论 #24287992 未加载
arey_abhishekover 4 years ago
Great product! How is the data corpus being fed? Could this work as a chrome extension for any page?
zitterbewegungover 4 years ago
I tried using tfjs and Bert with 87 kilobytes of text do you have a similar issue ?
评论 #24275081 未加载
nlouiover 4 years ago
Very cool. Also just signed up for your product beta.
scott31over 4 years ago
Should have been called Ctrl-Shift-F
xzyaoiover 4 years ago
Awesome!