Don't just read about AI, meet the builders. Integrating real-world insights is key to effective OpenClaw agent design. Here's how I use event data to refine agent focus: 1. Identify relevant events. Sites like TechCrunch list industry gatherings. Note speakers, topics. 2. Create a 'SpeakerProfile' OpenClaw agent. Feed it speaker names. Prompt: "Summarize this person's expertise and predict their likely stance on AI adoption in venture capital." 3. Build a 'TopicModel' agent. Feed it event topics. Prompt: "Create a topic model of venture capital investing in AI infrastructure. Identify key sub-themes and potential challenges." 4. Design an 'OpportunityDetector' agent. Input: outputs from SpeakerProfile and TopicModel. Prompt: "Based on these speaker profiles and topic models, identify potential investment opportunities and strategic partnerships for AI companies." 5. Run the agents daily, updating event info. Prioritize signals from events featuring TDK Ventures, Replit – they're driving the conversation. This keeps your OpenClaw agents ahead of the curve. Are you tuning your agents to real-world events?

BERTとTopicModelの融合「BERTopic」とは【実践編】 | soda | データ利活用・分析・AI開発 nico-soda.jp/blog/post/20230…

Under the hood update: Building on an earlier update, the scoring system will now further help to reduce flips on notes that are related to a topic that has an associated topic scoring model. If a note has a promising score, and is being scored by a TopicModel, but that TopicModel does not yet have enough ratings to produce a confident score, the note will show to Community Notes contributors as a proposed note preview for longer. This allows it to gather more ratings — reducing the chance of a flip — before it starts showing to everyone. As always, the code is open source: communitynotes.x.com/guide/e…

Glad to see you making use of the open source data & code. A key thing to know about topic models is that they score many notes that end up rated Helpful and shown across X, but the only time you'll see "TopicModel" show up as the "scored by" model is when the note is receiving a Needs More Ratings status. That's why you're not seeing the TopicModels assign a Helpful status, despite many notes they score ending up as Helpful -- the "scored by" field for those Helpful notes will list the initial model that assigned that Helpful status, e.g. CoreModel. Obviously there are a large number of Community Notes related to these topics that show widely across X. Topic models just help increase the likelihood that the notes will be found helpful to people from different points of view. They've been live for ~7 months and you can see full detail on how they work here: communitynotes.x.com/guide/e… Now, I know it can be frustrating if a note one feels is good does not get rated helpful and shown, or flips status. It does not mean the note is bad or flawed. There are undoubtedly notes that would be found broadly helpful but don't end up getting shown. We're continually improving the open source scoring algorithm to show as many helpful notes as possible. In general, the scoring system is designed to err on the side of ensuring the notes it shows are helpful and informative to people of different points of view, and avoiding showing notes that are not.

Ye, I hasn't been following this as closely in the last few months, and I remember that the acceptance rates for the Expansion model were considerably lower the last time I checked, but the stats for this TopicModel are just ridiculous.

TopicModel & model switching aren't actually new. Seen notes change from Core to Expansion, and Core to TM02. But seems to be happening more in recent months than before. Something changed under the hood, regardless of platform manipulation done by some of the worst offenders.

It's not an accident. TopicModel gets assigned to notes that get downvoted enough to not have a chance of turning "helpful". And I've seen it happen with multiple notes that went live to some of the worst offenders, who then sicced their followers on the notes to bring them down.

it will soon be available in journal version, but here's the preprint! it's a #bibliometric analysis of #transparency literature, mixed with a #topicmodel approach in order to identify trends in the sampled works it also criticizes the structures of knowledge production

Nice to see our recent work published in its final form @CellGenomics. Excited to work on more #topicmodel for #singlecell #CancerResearch

BERTとTopicModelの融合「BERTopic」とは nico-soda.jp/blog/post/20230…

BERTとTopicModelの融合「BERTopic」とは nico-soda.jp/blog/post/20230… この手法面白いな。使いたい

【note】TopicModelの最終形態？ Structured Topic Modelのご紹介 を公開しました！ ※こちらの記事は、2020年2月7日にRetrieva TECH BLOGにて掲載された記事を再掲載したものとなります。 note.com/retrieva/n/n430082f…

You can try the #KNIME Topic Explorer View component directly in your web browser as a #DataApp. In this case the view is executed on the output of an LDA #TopicModel on the 'CMU 2008 Political Blog Corpus'

BERTとTopicModelの融合「BERTopic」とは【実践編】 nico-soda.jp/blog/post/20230…

Excited to share our #ACL2023NLP paper which proposes a novel dynamic Brand-Topic Model, tracking brand scores and polarity-bearing topics over time. Join us at Poster Session 1. @yan_hanqi @yulanhe #NLProc #sentimentanalysis #TopicModel arxiv.org/abs/2301.07183.

For those interested in computational social science, this study by Miner et al. provides valuable insights on tradeoffs between topic models and human-generated coding in the context of workplace discrimination among physician mothers ➡️ buff.ly/3HaCQBN #topicmodel #NLP

🟡 Spotlight from #AAAI23: On Saturday, #Lamarr's Kostadin Cvejoski (@FraunhoferIAIS) highlighted work on a novel neural approach to the dynamic focused #TopicModel that outperforms state-of-the-art models in generalization tasks & converges two times faster in prediction tasks.

Exciting new paper by Miner and colleagues formally compares topic models to human-generated qualitative coding in the context of physician mothers' experiences of workplace discrimination. Check it out: buff.ly/3HaCQBN #topicmodel #NLP #qualitative

🚨New pub alert🚨 Excited to share this article on Topic Modeling! So lucky to have the chance to participate in this excellent project and collaborate with @l_lululy , Sei-Hill Kim, and Chang Won Choi! #topicmodel #CSS

New paper: Victor Bystrov, Viktoriia Naboka, Anna Staszewska-Bystrova und Peter Winker on "Cross-Corpora Comparisons of Topics and Topic Trends" #topicmodel #textanalysis degruyter.com/document/doi/1…