Asia Social Listening | LIHKG, PTT, HardwareZone & Telegram Brand Mention Aggregator

Give it one brand keyword and it scrapes related discussion across four platforms in a single run — Hong Kong's LIHKG, Taiwan's PTT, Singapore's HardwareZone, and Telegram — with automatic cross-platform deduplication, lightweight sentiment analysis, and competitor share-of-voice comparison. It hits each platform's public endpoints directly (it does not call our other actors as sub-actors, so you're never billed twice through a child actor's own metering), and outputs one flat, structured mention list.

What this is (and isn't)

• This is a meta-actor: it uses lightweight HTTP (got-scraping) to hit each platform's public endpoints directly, reusing (via copy, not import) the parsing logic from our own LIHKG / PTT / HardwareZone / Telegram scrapers. It does not invoke a separately-billed child actor.
• LIHKG runs a true, site-wide search. PTT, HardwareZone, and Telegram do keyword/title filtering within a target list you supply (boards / forums / channels) — not a site-wide search. Filter-based platforms can't surface discussion outside the list you give them; see Notes below.
• Threads (Asia · true search) is planned for v1.1 — it needs a browser (no public JSON endpoint), and the $0.05/mention price point can absorb that cost, but v1 ships with these four pure-HTTP platforms first.

Output Fields (MentionRow)

Field	Type	Description
mention_id	string	{platform}:{original ID}, stable across runs — used by deltaMode to diff new vs. seen
platforms	array	Which source it came from: lihkg / ptt / hardwarezone / telegram (all on by default)
brand_matched	string	The brand term matched (the primary brandKeyword, or one of the competitorBrands)
is_competitor	boolean	Whether this is a competitor mention (false = the primary brand itself)
matched_alias	string | null	The alias term matched (from searchAliases); null if matched on the primary term
text	string | null	Post title / body / message text
author	string | null	Author / poster
url	string | null	Link to the original post
created_at	string | null	Post time (ISO 8601; native precision varies by platform)
engagement_likes	number | null	Like count (most platforms don't expose this — reported honestly as null)
engagement_replies	number | null	Reply/comment count
engagement_views	number | null	View count
sentiment	string | null	positive / neutral / negative; null when sentiment analysis is off
sentiment_score	number | null	-1 (negative) to 1 (positive); null when sentiment analysis is off
is_duplicate	boolean	Whether this is a duplicate mention (same post, or a cross-platform repost); only false ones are billed
duplicate_of	string | null	Points to the first mention_id it duplicates, otherwise null
source_query	string	The query/filter term actually matched (primary term or alias)
scraped_at	string	Scrape timestamp (ISO 8601)

Platform Capabilities

Platform	Keyword Capability	Needs a Target List?
LIHKG (Hong Kong)	🟢 True, site-wide search	No
PTT (Taiwan)	🟡 Title filter within given boards	Yes (pttBoards, defaults to Gossiping / Stock)
HardwareZone (Singapore)	🟡 Title filter within given forums	Yes (hwzForums, defaults to 16 / EDMW)
Telegram	🟡 Text filter within given channels	Yes (telegramChannels; leaving it empty skips Telegram with a warning)

Input Fields

Field	Type	Default	Description
brandKeyword	string	—	The brand/keyword to monitor (required)
competitorBrands	array	[]	Competitor brands — each gets its own pass, letting you compare share of voice in the output
searchAliases	array	[]	Extra search terms for the primary brand (English name/abbreviation/common misspellings) — primary brand only
platforms	array	all four	Which platforms to monitor, see "Platform Capabilities" above
pttBoards	array	["Gossiping","Stock"]	Applies when platforms includes ptt
hwzForums	array	["16"]	Applies when platforms includes hardwarezone
telegramChannels	array	[]	Required when platforms includes telegram, otherwise skipped with a warning
lookbackDays	integer	7	Only scrape content from the last N days, range 1-90
maxMentionsPerPlatform	integer	100	Max mentions per platform per brand (after merging aliases), range 1-1000
sentimentAnalysis	boolean	true	Whether to run sentiment analysis (lightweight keyword wordlist, not AI)
deduplication	boolean	true	Whether to dedupe within and across platforms
deltaMode	boolean	false	Whether to return only mentions new since the last run
proxyConfiguration	proxy	Apify datacenter proxy	On by default, to avoid rate limiting from sharing egress IPs with other runs

Input Examples

Single brand, LIHKG and PTT only

{
  "brandKeyword": "誠品",
  "platforms": ["lihkg", "ptt"],
  "lookbackDays": 30,
  "maxMentionsPerPlatform": 40
}

Brand vs. competitor, with PTT/HardwareZone target lists

{
  "brandKeyword": "麥當勞",
  "competitorBrands": ["肯德基"],
  "platforms": ["lihkg", "ptt", "hardwarezone"],
  "pttBoards": ["Gossiping"],
  "hwzForums": ["16"],
  "lookbackDays": 30,
  "maxMentionsPerPlatform": 30
}

Scheduled Brand Monitoring

lookbackDays paired with deltaMode is designed for Apify Schedule:

• First run: deltaMode: false (or omit it) to get a full historical baseline.
• Subsequent scheduled runs (e.g. daily/weekly): turn deltaMode on, and set lookbackDays a bit longer than your schedule interval as a buffer (e.g. daily runs with lookbackDays: 3). Each run then returns only mentions not seen before (tracked via a brand-specific named KV store, SEEN-{hash of brand}, recording seen mention_ids across runs) — so a missed run won't drop data, and you won't see the same old mentions twice.
• Mentions already seen but filtered out this run by deltaMode still count toward skipped_delta in the run summary, so you can confirm the schedule is actually working (the number won't just sit at 0).
• Every run also writes a SUMMARY record to the KV store (counts per platform, sentiment breakdown, share-of-voice comparison) — handy for feeding a dashboard or pulling out for trend charts.

Notes

• Sentiment analysis is a lightweight, rule-based keyword wordlist (10 positive/negative terms each in Traditional Chinese and English), not an AI/LLM model. Score = (positive hits − negative hits) / (positive + negative + 1), with a ±0.15 threshold splitting positive/neutral/negative. Good for a quick read on general sentiment, not a substitute for precise sentiment judgment.
• Only public content is scraped — no account or login required. PTT's 18+ boards are handled automatically via the over18 cookie; every other platform here is a fully public page.
• True search vs. filter-based platforms: LIHKG runs an actual site-wide search, so in principle it can surface any matching discussion anywhere on the site. PTT/HardwareZone/Telegram only filter by title/text keyword within the boards/forums/channels you list — anything outside that list, or an unconfigured channel, is invisible to this actor. If your brand is discussed elsewhere, add that board/channel to the list yourself.
• Threads is planned for v1.1: Threads search is browser-rendered with no public JSON endpoint, so it needs a browser. This actor's v1 sticks to pure HTTP across the four platforms above. The $0.05/mention price point can actually absorb Threads' browser compute cost — v1.1 will add it via a separate browser adapter.
• searchAliases applies to the primary brand only: competitor brands are always searched/filtered using just their own brand term, never the primary brand's alias list.
• Deduplication rules: within a platform, dedup is by original post ID (a post counts once); across platforms, it's by a hash of normalized text (URLs/punctuation/emoji/whitespace stripped, lowercased) to catch reposts/duplicate content across platforms — the first one found is kept as canonical, and later ones are flagged is_duplicate: true. Only non-duplicate mentions are billed.
• maxMentionsPerPlatform is a per-platform, per-brand cap: if the primary brand has searchAliases set, the cap applies after merging and deduping the primary term with its aliases — it's not a separate cap per alias.
• If HardwareZone / PTT / LIHKG redesigns or adds anti-scraping measures down the line, that platform returns 0 rows with a warning instead of failing the whole run — the other platforms keep collecting data as normal.

FAQ

Q: Why are some fields (engagement_likes, etc.) always null?

A: Not every platform exposes like counts publicly. A null there isn't a broken field — that platform's public page simply doesn't carry that information.

Q: Why did I get far fewer mentions than expected?

A: The most common cause is that PTT/HardwareZone/Telegram filter within your target list, not a site-wide search — check that pttBoards/hwzForums/telegramChannels actually cover where your brand is likely to be discussed. It could also be lookbackDays set too short, or deltaMode being on with most mentions already seen in a prior run.

Q: Why do is_duplicate posts still show up in the dataset?

A: So you can see the full raw data and the duplicate relationship (you can trace duplicate_of back to the original mention). Duplicates are still pushed to the dataset — they're just not billed.

Q: Can I get only competitor mentions, without the primary brand?

A: Not currently — brandKeyword always runs. If you only want to compare competitorBrands against each other, add multiple competitors and just ignore rows where brand_matched equals your brandKeyword.

Disclaimer

This tool is for academic research and lawful business analysis only. Users must comply with the terms of service of each platform (LIHKG, PTT, HardwareZone, Telegram) and applicable laws. Do not use this tool to invade others' privacy, spread false information, or engage in any other unlawful activity. Sentiment analysis results are produced by a lightweight, rule-based algorithm, are for reference only, and do not constitute professional advice. The author accepts no responsibility for any misuse.

---

social listening, brand monitoring, LIHKG, PTT, HardwareZone, Telegram, sentiment analysis, share of voice, Asia social media monitoring