Near-duplicate applicant detection (collapse source reposts)

Exact ContentHash dedup misses the same ad reposted with slightly different text
(e.g. the ~18 repeated «کمک‌یار آقا»). DedupeTalentAsync collapses open aggregated
applicants by two high-precision signals — identical phone, or identical
(role, city, normalized description core with digits/«… پیش» time-phrases
stripped) — keeping the newest of each group. Runs at the end of both RunAsync
and ReprocessAsync; removed count surfaces in the run log.

Improvement 1 of the data-quality/SEO backlog.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

src/JobsMedical.Web/Services/Scraping/IngestionService.cs

@@ -147,6 +147,8 @@ public class IngestionService
 
         var summary = new IngestionSummary(results);
 
+        await DedupeTalentAsync(ct);   // collapse same-ad reposts the exact-hash dedup can't catch
+
         // Persist a run-log row so admins get a crawl history (with a per-source breakdown).
         if (results.Count > 0)
         {
@@ -246,19 +248,59 @@ public class IngestionService
         }
         await _db.SaveChangesAsync(ct);
 
+        var deduped = await DedupeTalentAsync(ct);   // collapse reposts the exact-hash dedup missed
+
         _db.IngestionRuns.Add(new IngestionRun
         {
-            Fetched = fetched, Queued = queued, Published = published, Flagged = flagged, Spam = spam, Duplicates = 0,
+            Fetched = fetched, Queued = queued, Published = published, Flagged = flagged, Spam = spam, Duplicates = deduped,
-            Detail = $"پردازش مجدد آیتم‌های ذخیره‌شده — {fetched} آیتم: {published} منتشر، {queued} صف، {flagged} پرچم، {spam} ردشده/قدیمی",
+            Detail = $"پردازش مجدد آیتم‌های ذخیره‌شده — {fetched} آیتم: {published} منتشر، {queued} صف، {flagged} پرچم، {spam} ردشده/قدیمی، {deduped} تکراریِ حذف‌شده",
         });
         await _db.SaveChangesAsync(ct);
-        _log.LogInformation("Reprocess done: items={F} published={P} queued={Q} flagged={Fl} discarded={S}",
+        _log.LogInformation("Reprocess done: items={F} published={P} queued={Q} flagged={Fl} discarded={S} deduped={D}",
-            fetched, published, queued, flagged, spam);
+            fetched, published, queued, flagged, spam, deduped);
 
         return new IngestionSummary(new List<SourceResult>
-            { new("پردازش مجدد", fetched, queued, published, flagged, spam, 0) });
+            { new("پردازش مجدد", fetched, queued, published, flagged, spam, deduped) });
     }
 
+    /// <summary>
+    /// Collapse near-duplicate aggregated APPLICANTS left when a source reposts the same ad (different
+    /// text → different ContentHash, so exact dedup missed them). Two high-precision signals: an
+    /// identical phone, or identical (role, city, normalized description core with digits/«… پیش»
+    /// time-phrases removed). Keeps the NEWEST of each group, deletes the rest. Returns the count removed.
+    /// </summary>
+    public async Task<int> DedupeTalentAsync(CancellationToken ct = default)
+    {
+        var rows = await _db.TalentListings
+            .Where(t => t.Status == ShiftStatus.Open && t.Source == ShiftSource.Aggregated)
+            .Select(t => new { t.Id, t.Phone, t.RoleId, t.CityId, t.Description, t.CreatedAt })
+            .ToListAsync(ct);
+
+        string? Sig(string? phone, int roleId, int cityId, string? desc)
+        {
+            var p = DigitsOnly(phone ?? "");
+            if (p.Length >= 7) return "p:" + p;                              // same number = same person/repost
+            var core = NormalizeFa(Regex.Replace(desc ?? "",
+                @"[0-9۰-۹]+|روز پیش|ساعت پیش|هفته پیش|دقیقه پیش|دیروز|پریروز", " ")).Trim();
+            if (core.Length < 15) return null;                              // too little to call it a dup safely
+            return $"t:{roleId}:{cityId}:{(core.Length > 100 ? core[..100] : core)}";
+        }
+
+        var toRemove = rows
+            .Select(r => new { r.Id, r.CreatedAt, Key = Sig(r.Phone, r.RoleId, r.CityId, r.Description) })
+            .Where(x => x.Key is not null)
+            .GroupBy(x => x.Key)
+            .SelectMany(g => g.OrderByDescending(x => x.CreatedAt).Skip(1).Select(x => x.Id))
+            .ToList();
+
+        if (toRemove.Count == 0) return 0;
+        var removed = await _db.TalentListings.Where(t => toRemove.Contains(t.Id)).ExecuteDeleteAsync(ct);
+        _log.LogInformation("Deduped {N} near-duplicate applicants.", removed);
+        return removed;
+    }
+
+    private static string DigitsOnly(string s) => new(HtmlUtil.ToLatinDigits(s).Where(char.IsDigit).ToArray());
+
     private static (RawListingStatus status, string? reason, int confidence) Decide(
         AppSetting s, ValidationResult val, AiAuditResult? ai)
     {