<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Problem Solving on Subculture Science Research Group&#39;s Blog</title>
    <link>https://www.subcul-science.com/en/tags/problem-solving/</link>
    <description>Recent content in Problem Solving on Subculture Science Research Group&#39;s Blog</description>
    <generator>Hugo</generator>
    <language>en-US</language>
    <lastBuildDate>Sun, 15 Nov 2026 15:30:00 +0900</lastBuildDate>
    <atom:link href="https://www.subcul-science.com/en/tags/problem-solving/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>The Small LLM That Answers Before It Thinks: Four Roles and a Comparison Design</title>
      <link>https://www.subcul-science.com/en/post/20261115umigame-gan-solution/</link>
      <pubDate>Sun, 15 Nov 2026 15:30:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115umigame-gan-solution/</guid>
      <description>Small LLMs commit to a hunch and stop tracking candidates. This post breaks down the failure modes, describes a loop that grows a creator and a solver in turn, and reports how far a per-turn reward actually moved the numbers.</description>
    </item>
    <item>
      <title>When the Timeline Fills Up with Reposts and Copies: Designing a Filter That Only Changes What You See</title>
      <link>https://www.subcul-science.com/en/post/20261115x-original-filter-solution/</link>
      <pubDate>Sun, 15 Nov 2026 15:10:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115x-original-filter-solution/</guid>
      <description>How to fix a timeline buried under reposts and re-uploaded images without deleting anything: why mute and keyword filters fall short, how to pick the original by timestamp, and how to keep a way back when the filter guesses wrong.</description>
    </item>
    <item>
      <title>Folding the Search for a Repost&#39;s Origin into One Query</title>
      <link>https://www.subcul-science.com/en/post/20261115origin-finder-solution/</link>
      <pubDate>Sun, 15 Nov 2026 14:30:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115origin-finder-solution/</guid>
      <description>Tracing a reposted image back to its original post is a manual loop across several reverse image search engines and dozens of page dates. How Origin Finder collapses that loop, and what it still leaves to a human.</description>
    </item>
    <item>
      <title>Turning One Image into a Talking 3D Character: Making the Build Repeatable</title>
      <link>https://www.subcul-science.com/en/post/202611153d-chatting-solution/</link>
      <pubDate>Sun, 15 Nov 2026 13:50:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/202611153d-chatting-solution/</guid>
      <description>Full mesh generation gives a different character on every run, and hand editing leaves no record. How driving VRoid Studio through an MCP turned character work into a repeatable cycle of change, export, compare, and accept.</description>
    </item>
    <item>
      <title>Reading Trend Performance in Numbers — Filling the Blind Spot in View-Count Rankings</title>
      <link>https://www.subcul-science.com/en/post/20261115youtube-analyser-solution/</link>
      <pubDate>Sun, 15 Nov 2026 13:30:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115youtube-analyser-solution/</guid>
      <description>A view-count ranking cannot tell you whether a trend video worked. This is how one app puts a viral coefficient, a median, and an engagement rate on one screen, aligns the comparison group, and where those numbers still fail.</description>
    </item>
    <item>
      <title>Stop Calculating Business Trip Travel Time by Hand</title>
      <link>https://www.subcul-science.com/en/post/20261115booking-butler-solution/</link>
      <pubDate>Sun, 15 Nov 2026 13:10:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115booking-butler-solution/</guid>
      <description>Copying addresses into a map app and subtracting travel time from every meeting is a nightly ritual for anyone who travels for work. BookingButler derives the origin from the calendar and works backwards from arrival. Here is what got automated and what stayed manual.</description>
    </item>
    <item>
      <title>A Search-Enabled Chat UI in the Shortest Path: A Research Assistant with Streamlit and a ReAct Agent</title>
      <link>https://www.subcul-science.com/en/post/20261115chatgpt-streamlit-solution/</link>
      <pubDate>Sun, 15 Nov 2026 12:50:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115chatgpt-streamlit-solution/</guid>
      <description>Ask a chat model about last week and you get stale knowledge. Then you search in another tab, paste the result, and ask again. Here is how a LangChain ReAct agent with Google and Wikipedia tools collapses that loop into one Streamlit file, and where the approach stops working.</description>
    </item>
    <item>
      <title>Cutting the Cost of Knowing What Is Trending on X</title>
      <link>https://www.subcul-science.com/en/post/20261115topic-digestx-solution/</link>
      <pubDate>Sun, 15 Nov 2026 12:30:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115topic-digestx-solution/</guid>
      <description>Following an X timeline costs three things: finding the right posts, reading them, and summarizing them. How TopicDigestX moves each to a server, the trade-offs, and what is still not built.</description>
    </item>
    <item>
      <title>Fine-tuning a Japanese LLM Without a GPU: One Mac, One Pipeline</title>
      <link>https://www.subcul-science.com/en/post/20261115ft-llm-solution/</link>
      <pubDate>Sun, 15 Nov 2026 11:50:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115ft-llm-solution/</guid>
      <description>No CUDA GPU, no cloud quota, no separate evaluation scripts. How FT_LLM turns fine-tuning a Japanese LLM into one Mac workflow, and where the approach still falls short.</description>
    </item>
    <item>
      <title>Three Constraints in Setlist Planning, and How Each One Was Removed</title>
      <link>https://www.subcul-science.com/en/post/20261115setlist-automation-solution/</link>
      <pubDate>Sun, 15 Nov 2026 11:30:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115setlist-automation-solution/</guid>
      <description>Planning a setlist where members switch instruments runs into three constraints: transition cost, stage time, and where the break goes. How the manual limits were framed, and how OR-Tools TSP plus a break inserter removed them.</description>
    </item>
    <item>
      <title>Reading Aloud in My Own Voice: Voice-Clone TTS from 30 Seconds of Audio</title>
      <link>https://www.subcul-science.com/en/post/20261115tts-solution/</link>
      <pubDate>Sun, 15 Nov 2026 11:10:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115tts-solution/</guid>
      <description>Ready-made TTS reads text well but never in your voice. Here is how a reference clip of about 30 seconds becomes a registered voice you can call over HTTP, using TADA and FastAPI.</description>
    </item>
    <item>
      <title>Filling a Quiet Discord Channel and a Broken News Flow with One Resident Bot</title>
      <link>https://www.subcul-science.com/en/post/20261115discord-bot-solution/</link>
      <pubDate>Sun, 15 Nov 2026 10:50:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115discord-bot-solution/</guid>
      <description>Two common problems in a small Discord server: chat that never gets going, and news that stays inside everyone&amp;rsquo;s own RSS reader. How one resident bot addresses both with a persona, scheduled posts, and interests that grow from conversation.</description>
    </item>
    <item>
      <title>No One to Critique Your Drawing, and No Wish to Upload It</title>
      <link>https://www.subcul-science.com/en/post/20261115dessan-sensei-solution/</link>
      <pubDate>Sun, 15 Nov 2026 10:30:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115dessan-sensei-solution/</guid>
      <description>Drawing alone means no reviewer nearby, and uploading practice sketches to a cloud service feels wrong. How Dessan-sensei removed both walls with a local vision model on the LAN, using text-only history as context and level-aware prompting.</description>
    </item>
    <item>
      <title>Automating the Split of Stream Archives: Three Walls of Manual Work, Memory, and Speed</title>
      <link>https://www.subcul-science.com/en/post/20261115jam-slicer-solution/</link>
      <pubDate>Sun, 15 Nov 2026 10:10:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115jam-slicer-solution/</guid>
      <description>Splitting a long stream archive into individual songs runs into three walls: manual timestamps, huge memory use, and waiting. How JamSlicer removes each one, with the design and code behind it.</description>
    </item>
    <item>
      <title>A LINE Butler That Keeps Its Character While Recommending Products</title>
      <link>https://www.subcul-science.com/en/post/20261115turtle-buttler-solution/</link>
      <pubDate>Sun, 15 Nov 2026 09:50:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115turtle-buttler-solution/</guid>
      <description>A characterful bot turns into a salesperson when you add product links, and a sales bot loses the character. How turtle-buttler split voice, catalogue, and judgement across a prompt, an API, and the model.</description>
    </item>
    <item>
      <title>When Your Likes Get Buried: Designing Capture and Rediscovery for X Favorites</title>
      <link>https://www.subcul-science.com/en/post/20261115x-fav-gallery-solution/</link>
      <pubDate>Sun, 15 Nov 2026 09:30:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115x-fav-gallery-solution/</guid>
      <description>Likes on X sink into the timeline the moment you tap them, and there is no way to search by what an image looks like. How Refav Gallery split capture, storage, and rediscovery to fix that, including the tradeoffs and measured results.</description>
    </item>
    <item>
      <title>Preventing Missed Sequels by Design: Stop Relying on Remembering at the Bookstore</title>
      <link>https://www.subcul-science.com/en/post/20261115tsuzukan-solution/</link>
      <pubDate>Sun, 15 Nov 2026 09:10:00 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20261115tsuzukan-solution/</guid>
      <description>Missing the next volume is not a memory problem, it is a problem of where the state lives. How tsuzukan records owned volumes by ISBN and reports what the series is missing, including the cases a machine cannot decide.</description>
    </item>
    <item>
      <title>How to Solve Audio File Volume Inconsistency and Quality Unification Issues with ffmpeg Normalization</title>
      <link>https://www.subcul-science.com/en/post/20250705audio-volume-normalization-solution/</link>
      <pubDate>Sat, 05 Jul 2025 21:05:51 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20250705audio-volume-normalization-solution/</guid>
      <description>&lt;h2 id=&#34;audio-volume-issues-in-audio-file-processing&#34;&gt;Audio Volume Issues in Audio File Processing&lt;/h2&gt;&#xA;&lt;p&gt;When producing and distributing audio content, do you face these problems?&lt;/p&gt;&#xA;&lt;h3 id=&#34;1-volume-inconsistency-issues&#34;&gt;1. Volume Inconsistency Issues&lt;/h3&gt;&#xA;&lt;ul&gt;&#xA;&lt;li&gt;&lt;strong&gt;Volume levels are not unified across multiple audio files&lt;/strong&gt;&lt;/li&gt;&#xA;&lt;li&gt;Volume differences occur due to different recording environments and equipment&lt;/li&gt;&#xA;&lt;li&gt;Listeners need to frequently adjust volume levels&lt;/li&gt;&#xA;&lt;/ul&gt;&#xA;&lt;h3 id=&#34;2-quality-inconsistency-issues&#34;&gt;2. Quality Inconsistency Issues&lt;/h3&gt;&#xA;&lt;ul&gt;&#xA;&lt;li&gt;&lt;strong&gt;Noise and unwanted frequencies are mixed in&lt;/strong&gt;&lt;/li&gt;&#xA;&lt;li&gt;Silent sections are too long and difficult to listen to&lt;/li&gt;&#xA;&lt;li&gt;Unable to achieve professional-quality audio&lt;/li&gt;&#xA;&lt;/ul&gt;&#xA;&lt;h3 id=&#34;3-manual-processing-limitations&#34;&gt;3. Manual Processing Limitations&lt;/h3&gt;&#xA;&lt;ul&gt;&#xA;&lt;li&gt;&lt;strong&gt;Processing large numbers of audio files individually is inefficient&lt;/strong&gt;&lt;/li&gt;&#xA;&lt;li&gt;Automation is difficult with GUI audio editing software&lt;/li&gt;&#xA;&lt;li&gt;Applying consistent processing standards is challenging&lt;/li&gt;&#xA;&lt;/ul&gt;&#xA;&lt;h2 id=&#34;real-world-audio-quality-challenge-cases&#34;&gt;Real-world Audio Quality Challenge Cases&lt;/h2&gt;&#xA;&lt;h3 id=&#34;failure-case-limitations-of-manual-adjustment&#34;&gt;Failure Case: Limitations of Manual Adjustment&lt;/h3&gt;&#xA;&lt;div class=&#34;highlight&#34;&gt;&lt;pre tabindex=&#34;0&#34; class=&#34;chroma&#34;&gt;&lt;code class=&#34;language-bash&#34; data-lang=&#34;bash&#34;&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# Traditional approach&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# 1. Open each file in audio editing software&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# 2. Visually and auditorily adjust levels&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# 3. Manually apply noise reduction&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# 4. Manually cut silent sections&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# Problems:&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# - Time-consuming for processing large numbers of files&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# - Processing standards are subjective and inconsistent&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# - Quality variations due to human errors&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The solution to this problem is &lt;strong&gt;automated volume normalization with ffmpeg&lt;/strong&gt;.&lt;/p&gt;</description>
    </item>
    <item>
      <title>How to Efficiently Solve Low Accuracy and High Cost Issues in Japanese Text Generation with T5</title>
      <link>https://www.subcul-science.com/en/post/20250705japanese-text-generation-solution/</link>
      <pubDate>Sat, 05 Jul 2025 21:00:38 +0900</pubDate>
      <guid>https://www.subcul-science.com/en/post/20250705japanese-text-generation-solution/</guid>
      <description>&lt;h2 id=&#34;challenges-in-japanese-text-generation&#34;&gt;Challenges in Japanese Text Generation&lt;/h2&gt;&#xA;&lt;p&gt;When working on Japanese text summarization, title generation, and document classification tasks, do you face these problems?&lt;/p&gt;&#xA;&lt;h3 id=&#34;1-accuracy-issues&#34;&gt;1. Accuracy Issues&lt;/h3&gt;&#xA;&lt;ul&gt;&#xA;&lt;li&gt;&lt;strong&gt;Traditional rule-based methods cannot generate natural Japanese text&lt;/strong&gt;&lt;/li&gt;&#xA;&lt;li&gt;English-oriented models cannot handle Japanese grammar and expressions&lt;/li&gt;&#xA;&lt;li&gt;Need to build separate models for multiple tasks&lt;/li&gt;&#xA;&lt;/ul&gt;&#xA;&lt;h3 id=&#34;2-development-cost-issues&#34;&gt;2. Development Cost Issues&lt;/h3&gt;&#xA;&lt;ul&gt;&#xA;&lt;li&gt;&lt;strong&gt;Time and resources required for task-specific model development&lt;/strong&gt;&lt;/li&gt;&#xA;&lt;li&gt;Different approaches needed for document classification, summarization, and title generation&lt;/li&gt;&#xA;&lt;li&gt;Enormous effort required for preparing training data and building models&lt;/li&gt;&#xA;&lt;/ul&gt;&#xA;&lt;h3 id=&#34;3-operational-complexity&#34;&gt;3. Operational Complexity&lt;/h3&gt;&#xA;&lt;ul&gt;&#xA;&lt;li&gt;&lt;strong&gt;Need to manage and operate multiple models&lt;/strong&gt;&lt;/li&gt;&#xA;&lt;li&gt;Different APIs and interfaces for each task&lt;/li&gt;&#xA;&lt;li&gt;Complex model updates and maintenance&lt;/li&gt;&#xA;&lt;/ul&gt;&#xA;&lt;h2 id=&#34;real-world-text-generation-challenge-cases&#34;&gt;Real-world Text Generation Challenge Cases&lt;/h2&gt;&#xA;&lt;h3 id=&#34;failure-case-limitations-of-task-specific-individual-development&#34;&gt;Failure Case: Limitations of Task-specific Individual Development&lt;/h3&gt;&#xA;&lt;div class=&#34;highlight&#34;&gt;&lt;pre tabindex=&#34;0&#34; class=&#34;chroma&#34;&gt;&lt;code class=&#34;language-python&#34; data-lang=&#34;python&#34;&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# Traditional approach&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;n&#34;&gt;classification_model&lt;/span&gt; &lt;span class=&#34;o&#34;&gt;=&lt;/span&gt; &lt;span class=&#34;n&#34;&gt;load_bert_classifier&lt;/span&gt;&lt;span class=&#34;p&#34;&gt;()&lt;/span&gt;      &lt;span class=&#34;c1&#34;&gt;# For document classification&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;n&#34;&gt;summarization_model&lt;/span&gt; &lt;span class=&#34;o&#34;&gt;=&lt;/span&gt; &lt;span class=&#34;n&#34;&gt;load_summarization_model&lt;/span&gt;&lt;span class=&#34;p&#34;&gt;()&lt;/span&gt;   &lt;span class=&#34;c1&#34;&gt;# For summarization&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;n&#34;&gt;title_generation_model&lt;/span&gt; &lt;span class=&#34;o&#34;&gt;=&lt;/span&gt; &lt;span class=&#34;n&#34;&gt;load_title_model&lt;/span&gt;&lt;span class=&#34;p&#34;&gt;()&lt;/span&gt;        &lt;span class=&#34;c1&#34;&gt;# For title generation&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# Problems:&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# - Managing 3 separate models&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# - 3x memory usage&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;span class=&#34;line&#34;&gt;&lt;span class=&#34;cl&#34;&gt;&lt;span class=&#34;c1&#34;&gt;# - High development and maintenance costs&lt;/span&gt;&#xA;&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The solution to this problem is &lt;strong&gt;Japanese T5 (Text-To-Text Transfer Transformer)&lt;/strong&gt;.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
