soroush.asadi
1f628d971e
A generic hiring ad like «پرستار درمانگاه» was published as a dated SHIFT with an invented date («فردا») and default hours («۰۸:۰۰–۱۴:۰۰») the source never stated — because classification defaulted to Shift. Now a dated Shift is only produced when the text carries an explicit shift signal (شیفت/آنکال/کشیک/نوبت); everything else is an ongoing hiring post → Job (no date to invent). Fixed in both the parser default and the Publish branch (so an AI mislabel can''t force a shift either). ReclassifyMisclassifiedShiftsAsync (in the post-ingest auto-cleanup) converts the existing signal-less aggregated shifts into jobs in place — copies the content to a JobOpening and archives the old shift (its URL 410s). After one pass it''s a no-op since new ads no longer become shifts. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
451 lines
26 KiB
C#
451 lines
26 KiB
C#
using System.Text.RegularExpressions;
|
||
using JobsMedical.Web.Models;
|
||
|
||
namespace JobsMedical.Web.Services;
|
||
|
||
/// <summary>One contact channel pulled from a post (type + raw value).</summary>
|
||
public record ParsedContact(ContactType Type, string Value);
|
||
|
||
/// <summary>Structured guess extracted from a raw channel post. All fields are best-effort.</summary>
|
||
public class ParsedListing
|
||
{
|
||
public ListingKind Kind { get; set; } = ListingKind.Shift;
|
||
public string? RoleName { get; set; } // primary role (first match)
|
||
public List<string> RoleNames { get; set; } = new(); // all roles in the ad (e.g. سالمند + کودک)
|
||
public ShiftType? ShiftType { get; set; }
|
||
public EmploymentType? EmploymentType { get; set; }
|
||
public long? PayAmount { get; set; } // shift pay or single salary figure
|
||
public int? SharePercent { get; set; } // profit-share % (درصدی / سهم درآمد)
|
||
public bool PayNegotiable { get; set; }
|
||
public Gender Gender { get; set; } = Gender.Any; // جنسیت مورد نیاز
|
||
public string? CityName { get; set; }
|
||
public string? DistrictName { get; set; }
|
||
public string? FacilityName { get; set; } // hospital/clinic name guessed from the text
|
||
public string? Phone { get; set; }
|
||
|
||
// «آماده به کار» (talent) extras — populated when Kind == Talent.
|
||
public string? PersonName { get; set; } // «دکتر سپیده علیزاده»
|
||
public int? YearsExperience { get; set; } // سابقه (سال)
|
||
public bool IsLicensed { get; set; } // پروانهدار
|
||
public string? AreaNote { get; set; } // «فقط منطقه ۱»
|
||
public List<ParsedContact> Contacts { get; set; } = new(); // phones, email, socials…
|
||
public List<string> Tags { get; set; } = new(); // cert/skill keywords for search
|
||
public List<string> Notes { get; set; } = new(); // what was/wasn't detected (shown to admin)
|
||
}
|
||
|
||
/// <summary>
|
||
/// Turns a messy Persian channel/Divar post into a structured listing guess. This is the
|
||
/// Stage-1 implementation: transparent keyword + regex heuristics, no AI dependency (important
|
||
/// since LLM APIs are blocked from Iran). A future LlmListingParser can implement the same
|
||
/// interface and be swapped in via DI without touching the admin queue.
|
||
/// </summary>
|
||
public interface IListingParser
|
||
{
|
||
ParsedListing Parse(string rawText, IEnumerable<string> knownRoles,
|
||
IEnumerable<string> knownCities, IEnumerable<string> knownDistricts);
|
||
}
|
||
|
||
public class HeuristicListingParser : IListingParser
|
||
{
|
||
public ParsedListing Parse(string raw, IEnumerable<string> knownRoles,
|
||
IEnumerable<string> knownCities, IEnumerable<string> knownDistricts)
|
||
{
|
||
var p = new ParsedListing();
|
||
var text = Normalize(raw);
|
||
|
||
// --- Kind: talent (worker offers themselves) vs shift vs hiring ---
|
||
// Talent is checked first: «آماده به کار/همکاری», «جویای کار» mean the *person* is
|
||
// available — distinct from an employer's «دعوت به همکاری».
|
||
bool talentSignals = ContainsAny(text,
|
||
"آماده به کار", "آمادهبهکار", "آماده همکاری", "آمادهی همکاری", "آماده ی همکاری",
|
||
"آماده فعالیت", "جویای کار", "جویای کار هستم", "متقاضی کار", "نیازمند کار",
|
||
"آماده انجام", "میتوانم همکاری", "میتوانم همکاری", "حاضر به همکاری");
|
||
bool jobSignals = ContainsAny(text, "استخدام", "جذب", "دعوت به همکاری", "نیازمندیم", "نیازمند است", "حقوق ثابت");
|
||
bool shiftSignals = ContainsAny(text, "شیفت", "آنکال", "انکال", "نوبت", "کشیک");
|
||
if (talentSignals)
|
||
{
|
||
p.Kind = ListingKind.Talent;
|
||
p.Notes.Add("نوع: آماده به کار (تشخیص خودکار)");
|
||
}
|
||
else
|
||
{
|
||
// A dated SHIFT requires an explicit shift signal («شیفت/آنکال/کشیک/نوبت»). Otherwise the ad
|
||
// is an ongoing hiring post → Job. (Defaulting to Shift forced a fabricated date/time onto
|
||
// generic ads like «پرستار درمانگاه», which the source never stated.)
|
||
p.Kind = shiftSignals ? ListingKind.Shift : ListingKind.Job;
|
||
p.Notes.Add(p.Kind == ListingKind.Shift ? "نوع: شیفت (تشخیص خودکار)" : "نوع: استخدام (تشخیص خودکار)");
|
||
}
|
||
|
||
// --- Roles (an ad can name several at once: «پرستار سالمند و کودک و همراه بیمار») ---
|
||
var known = knownRoles.ToList();
|
||
var hits = new List<string>();
|
||
// Exact taxonomy matches (longest first so «پزشک متخصص» beats «پزشک»).
|
||
foreach (var role in known.OrderByDescending(r => r.Length))
|
||
if (text.Contains(Normalize(role))) hits.Add(role);
|
||
// Drop a role that's a substring of a longer matched role (پرستار ⊂ پرستار سالمندان).
|
||
hits = hits.Where(r => !hits.Any(o => o != r && o.Contains(r))).Distinct().ToList();
|
||
|
||
// Synonyms → canonical role names (covers terms not written verbatim). Only add a canonical
|
||
// that actually exists in the taxonomy, and isn't already a hit.
|
||
void AddSyn(string canonical, params string[] needles)
|
||
{
|
||
if (ContainsAny(text, needles) && known.Contains(canonical) && !hits.Contains(canonical))
|
||
hits.Add(canonical);
|
||
}
|
||
AddSyn("پرستار سالمندان", "سالمند", "سالمندان", "نگهداری سالمند");
|
||
AddSyn("دندانپزشک", "دندان", "دندانپزشک", "دندانپزشک");
|
||
AddSyn("تکنسین اتاق عمل", "اتاق عمل", "اسکراب");
|
||
AddSyn("تکنسین فوریتهای پزشکی", "فوریت", "اورژانس پیش بیمارستانی", "آمبولانس");
|
||
AddSyn("کارشناس آزمایشگاه", "آزمایشگاه", "علوم آزمایشگاهی", "نمونه گیر");
|
||
AddSyn("ماما", "مامایی");
|
||
AddSyn("پرستار", "بهیار", "کمک بهیار", "کمک پرستار", "بیماربر", "مراقب", "همراه بیمار",
|
||
"کودک", "اطفال", "نوزاد", "تزریقات", "پانسمان");
|
||
AddSyn("پزشک متخصص", "فوق تخصص", "متخصص");
|
||
AddSyn("پزشک عمومی", "پزشک", "دکتر", "طبیب");
|
||
|
||
p.RoleNames = hits.Distinct().Take(4).ToList(); // cap fan-out
|
||
p.RoleName = p.RoleNames.FirstOrDefault();
|
||
p.Notes.Add(p.RoleNames.Count == 0 ? "نقش: تشخیص داده نشد" : $"نقشها: {string.Join("، ", p.RoleNames)}");
|
||
|
||
// --- Shift type ---
|
||
if (ContainsAny(text, "آنکال", "انکال")) p.ShiftType = Models.ShiftType.OnCall;
|
||
else if (text.Contains("شب")) p.ShiftType = Models.ShiftType.Night;
|
||
else if (text.Contains("عصر")) p.ShiftType = Models.ShiftType.Evening;
|
||
else if (ContainsAny(text, "صبح", "روز")) p.ShiftType = Models.ShiftType.Day;
|
||
|
||
// --- Employment type ---
|
||
if (ContainsAny(text, "پاره وقت", "پارهوقت", "پارت تایم")) p.EmploymentType = Models.EmploymentType.PartTime;
|
||
else if (text.Contains("طرح")) p.EmploymentType = Models.EmploymentType.Plan;
|
||
else if (text.Contains("قرارداد")) p.EmploymentType = Models.EmploymentType.Contract;
|
||
else if (ContainsAny(text, "تمام وقت", "تماموقت")) p.EmploymentType = Models.EmploymentType.FullTime;
|
||
|
||
// --- Gender requirement ---
|
||
if (ContainsAny(text, "خانم", "خانوم", "بانو", "زن ", "مامای")) p.Gender = Gender.Female;
|
||
else if (ContainsAny(text, "آقا", "اقا", "مرد ", "مرد،", "پسر")) p.Gender = Gender.Male;
|
||
if (p.Gender != Gender.Any)
|
||
p.Notes.Add($"جنسیت: {(p.Gender == Gender.Female ? "خانم" : "آقا")}");
|
||
|
||
// --- City / district ---
|
||
p.CityName = knownCities.FirstOrDefault(c => text.Contains(Normalize(c)));
|
||
p.DistrictName = knownDistricts.OrderByDescending(d => d.Length)
|
||
.FirstOrDefault(d => text.Contains(Normalize(d)));
|
||
|
||
// --- Profit share (درصدی / سهم) ---
|
||
var latinForShare = ToLatinDigits(text);
|
||
var share = Regex.Match(latinForShare, @"(\d{1,3})\s*(?:٪|%|درصد)");
|
||
if (!share.Success) share = Regex.Match(latinForShare, @"(?:٪|%)\s*(\d{1,3})");
|
||
if (share.Success && int.TryParse(share.Groups[1].Value, out var pct) && pct is > 0 and <= 100)
|
||
{ p.SharePercent = pct; p.Notes.Add($"سهم درآمد: {pct}٪"); }
|
||
else if (ContainsAny(text, "درصدی", "سهم درآمد", "شراکت", "پورسانت"))
|
||
{ p.Notes.Add("پرداخت درصدی/سهمی (درصد نامشخص)"); }
|
||
|
||
// --- Fixed pay (strip phone numbers first so they're never read as money) ---
|
||
// A STATED amount wins over «توافقی»: ads often say a number AND «… بقیه توافقی»; showing the
|
||
// figure is far more useful than «توافقی». Fall back to negotiable only when no amount is found.
|
||
var amount = ExtractAmount(StripPhones(text));
|
||
if (amount is not null) { p.PayAmount = amount; p.Notes.Add($"حقوق تخمینی: {amount:#,0} تومان"); }
|
||
else if (ContainsAny(text, "توافقی", "توافق")) { p.PayNegotiable = true; p.Notes.Add("حقوق: توافقی"); }
|
||
else if (p.SharePercent is null) p.Notes.Add("حقوق: تشخیص داده نشد");
|
||
|
||
// --- Talent extras (only meaningful for «آماده به کار») ---
|
||
if (p.Kind == ListingKind.Talent)
|
||
{
|
||
var latinT = ToLatinDigits(text);
|
||
var exp = Regex.Match(latinT, @"سابقه[^\d]{0,8}(\d{1,2})\s*سال");
|
||
if (!exp.Success) exp = Regex.Match(latinT, @"(\d{1,2})\s*سال\s*سابقه");
|
||
if (exp.Success && int.TryParse(exp.Groups[1].Value, out var yrs) && yrs is > 0 and <= 60)
|
||
{ p.YearsExperience = yrs; p.Notes.Add($"سابقه: {yrs} سال"); }
|
||
|
||
p.IsLicensed = ContainsAny(text, "پروانه دار", "پروانهدار", "دارای پروانه", "پروانه فعالیت", "پروانه طبابت");
|
||
if (p.IsLicensed) p.Notes.Add("پروانهدار");
|
||
|
||
p.PersonName = ExtractPersonName(text);
|
||
if (p.PersonName is not null) p.Notes.Add($"نام: {p.PersonName}");
|
||
|
||
var area = Regex.Match(text, @"منطقه\s*[۰-۹0-9]{1,2}");
|
||
if (area.Success) { p.AreaNote = area.Value.Trim(); p.Notes.Add($"محدوده: {p.AreaNote}"); }
|
||
}
|
||
|
||
// --- Facility name (بیمارستان/درمانگاه/کلینیک ... + the distinctive name) ---
|
||
if (p.Kind != ListingKind.Talent)
|
||
{
|
||
p.FacilityName = ExtractFacilityName(text);
|
||
if (p.FacilityName is not null) p.Notes.Add($"مرکز: {p.FacilityName}");
|
||
}
|
||
|
||
// --- Tags (certs/skills for deep search): mmt, icu, پروانهدار, اتاق عمل … ---
|
||
p.Tags = ExtractTags(text);
|
||
if (p.RoleNames.Count > 0) p.Tags.AddRange(p.RoleNames);
|
||
if (p.IsLicensed && !p.Tags.Contains("پروانهدار")) p.Tags.Add("پروانهدار");
|
||
p.Tags = p.Tags.Distinct().ToList();
|
||
|
||
// --- Contacts (phones, email, socials — one ad may have several) ---
|
||
p.Contacts = ExtractContacts(raw ?? text);
|
||
p.Phone = p.Contacts.FirstOrDefault(c => c.Type is ContactType.Mobile or ContactType.Phone)?.Value;
|
||
if (p.Contacts.Count > 0)
|
||
p.Notes.Add("راههای ارتباطی: " + string.Join("، ", p.Contacts.Select(c => ContactLabel(c.Type))));
|
||
|
||
return p;
|
||
}
|
||
|
||
// Words that introduce a facility name, longest/most-specific first.
|
||
private static readonly string[] FacilityKeywords =
|
||
{
|
||
"بیمارستان", "زایشگاه", "پلی کلینیک", "پلیکلینیک", "درمانگاه", "کلینیک",
|
||
"مرکز درمانی", "مرکز جراحی", "مجتمع پزشکی", "مجتمع درمانی", "مرکز", "مجتمع",
|
||
"آزمایشگاه", "مطب", "خانه سالمندان", "سرای سالمندان",
|
||
};
|
||
|
||
// Words that clearly aren't part of a facility's name — stop collecting here.
|
||
private static readonly string[] NameStops =
|
||
{
|
||
"جهت", "برای", "به", "با", "در", "از", "که", "نیاز", "نیازمند", "استخدام", "جذب",
|
||
"دعوت", "همکاری", "واقع", "آدرس", "تلفن", "شماره", "شیفت", "ساعت", "حقوق", "روز",
|
||
"شب", "صبح", "عصر", "می", "ها", "این", "یک", "محترم",
|
||
};
|
||
|
||
/// <summary>Best-effort hospital/clinic name: a facility keyword plus up to three name words.</summary>
|
||
private static string? ExtractFacilityName(string text)
|
||
{
|
||
foreach (var kw in FacilityKeywords)
|
||
{
|
||
var idx = text.IndexOf(kw, StringComparison.Ordinal);
|
||
if (idx < 0) continue;
|
||
var after = text[(idx + kw.Length)..];
|
||
var words = after.Split(
|
||
new[] { ' ', '\n', '\r', '\t', '،', ',', '.', '؛', ':', '(', ')', '-', '/', '«', '»', '"' },
|
||
StringSplitOptions.RemoveEmptyEntries);
|
||
var picked = new List<string>();
|
||
foreach (var w in words)
|
||
{
|
||
if (NameStops.Contains(w)) break;
|
||
if (Regex.IsMatch(w, @"\d")) break; // numbers/phones aren't names
|
||
if (!w.Any(char.IsLetter)) break; // emoji / punctuation («📍») isn't a name
|
||
if (w.Length == 1) break; // stray letters
|
||
picked.Add(w);
|
||
if (picked.Count >= 3) break;
|
||
}
|
||
if (picked.Count == 0) continue; // bare keyword (e.g. just «بیمارستان») isn't useful
|
||
var candidate = (kw + " " + string.Join(" ", picked)).Trim();
|
||
// Reject names that are only filler/verb/source noise («بیمارستان هستم», «... از مدجابز») —
|
||
// a real name couldn't be extracted, so fall back to the shared placeholder downstream.
|
||
if (Scraping.FacilityMatcher.IsJunkName(candidate)) continue;
|
||
return candidate;
|
||
}
|
||
return null;
|
||
}
|
||
|
||
// Titles that introduce a person's name in «آماده به کار» posts.
|
||
private static readonly string[] PersonTitles = { "دکتر", "خانم دکتر", "آقای دکتر", "مهندس", "سرکار خانم", "جناب آقای", "خانم", "آقای" };
|
||
|
||
// Words that are NOT a person's name — verbs/fillers/availability/role words the extractor was
|
||
// grabbing after a title («خانم هستم»، «دکتر ام»، «دکتر داروساز آماده»). Stop collecting at one.
|
||
private static readonly string[] NameNoise =
|
||
{
|
||
"هستم", "هستیم", "هستش", "ام", "بودم", "میباشم", "میباشد", "باشم", "آماده", "آمادهام",
|
||
"جویای", "بکار", "بهکار", "کار", "همکاری", "نیازمند", "استخدام", "جذب", "عزیز", "محترم",
|
||
"گرامی", "خانم", "آقا", "اقا", "دکتر", "پزشک", "پرستار", "بهیار", "ماما", "دندانپزشک",
|
||
"داروساز", "تکنسین", "کارشناس", "متخصص", "عمومی", "مراقب", "کمک",
|
||
};
|
||
|
||
/// <summary>Best-effort person name: a title (دکتر/خانم/…) plus up to two following words.</summary>
|
||
private static string? ExtractPersonName(string text)
|
||
{
|
||
foreach (var title in PersonTitles)
|
||
{
|
||
var idx = text.IndexOf(title, StringComparison.Ordinal);
|
||
if (idx < 0) continue;
|
||
var after = text[(idx + title.Length)..];
|
||
var words = after.Split(
|
||
new[] { ' ', '\n', '\r', '\t', '،', ',', '.', '؛', ':', '(', ')', '-', '/' },
|
||
StringSplitOptions.RemoveEmptyEntries);
|
||
var picked = new List<string>();
|
||
foreach (var w in words)
|
||
{
|
||
if (NameStops.Contains(w)) break;
|
||
if (NameNoise.Any(n => Normalize(n) == Normalize(w))) break; // «خانم هستم»/«دکتر ام»…
|
||
if (Regex.IsMatch(w, @"[\d]")) break;
|
||
if (w.Length == 1) break;
|
||
picked.Add(w);
|
||
if (picked.Count >= 2) break;
|
||
}
|
||
if (picked.Count == 0) continue;
|
||
return (title + " " + string.Join(" ", picked)).Trim();
|
||
}
|
||
return null;
|
||
}
|
||
|
||
/// <summary>Remove phone numbers (and «شماره تماس…» lines) so they're not mistaken for money.</summary>
|
||
private static string StripPhones(string text)
|
||
{
|
||
var t = Regex.Replace(text, @"شماره\s*(?:تماس|موبایل|همراه|ثابت|تلفن)[^\n]*", " ");
|
||
t = ToLatinDigits(t);
|
||
t = Regex.Replace(t, @"(?<!\d)(?:\+?98|0)?9\d{9}(?!\d)", " "); // mobile
|
||
t = Regex.Replace(t, @"(?<!\d)0\d{2,3}[\s-]?\d{7,8}(?!\d)", " "); // landline
|
||
return t;
|
||
}
|
||
|
||
/// <summary>Pull a figure out of free text and normalize to TOMAN (ریال → تومان = ÷۱۰),
|
||
/// handling «میلیون» and Persian digits.</summary>
|
||
private static long? ExtractAmount(string text)
|
||
{
|
||
var latin = ToLatinDigits(text);
|
||
bool hasToman = latin.Contains("تومان") || latin.Contains("تومن");
|
||
bool hasRial = (latin.Contains("ریال") || latin.Contains("ريال")) && !hasToman;
|
||
|
||
// Iranian salary shorthand: a 1–3 digit number means MILLIONS of toman — «۱۵ تومان»،
|
||
// «۴۰ تا ۵۰ تومان»، «۲۰ میلیون»، «۲۰م». Take the LOWER bound of a range. The lookarounds keep
|
||
// this from ever matching part of a long literal-toman number (the digits must end at the unit).
|
||
var collo = Regex.Match(latin,
|
||
@"(?<!\d)(\d{1,3})(?:\s*تا\s*(\d{1,3}))?\s*(?:میلیون|م(?![ا-یA-Za-z])|تومان|تومن)(?!\s*\d)");
|
||
if (collo.Success && int.TryParse(collo.Groups[1].Value, out var lo) && lo is > 0 and <= 500)
|
||
return (long)lo * 1_000_000;
|
||
|
||
// e.g. "۲ میلیون" / "2.5 میلیون [ریال]"
|
||
var million = Regex.Match(latin, @"(\d+(?:[.,]\d+)?)\s*میلیون\s*(ریال|ريال)?");
|
||
if (million.Success && double.TryParse(million.Groups[1].Value.Replace(",", "."),
|
||
System.Globalization.NumberStyles.Any, System.Globalization.CultureInfo.InvariantCulture, out var m))
|
||
{
|
||
var val = (long)(m * 1_000_000);
|
||
if (million.Groups[2].Success) val /= 10; // «میلیون ریال»
|
||
return val;
|
||
}
|
||
|
||
// Largest plain number that looks like money (6–10 digits, no leading zero — a leading
|
||
// zero or 11+ digits means it's a phone/id). Convert ریال→تومان by the unit next to the
|
||
// number, else by the ad's overall currency.
|
||
long best = 0;
|
||
foreach (Match num in Regex.Matches(latin, @"(?<!\d)([1-9][\d٬,،.]{4,})\s*(ریال|ريال|تومان|تومن)?"))
|
||
{
|
||
var digits = Regex.Replace(num.Groups[1].Value, @"[^\d]", "");
|
||
if (digits.Length is < 6 or > 10 || !long.TryParse(digits, out var v)) continue;
|
||
var unit = num.Groups[2].Value;
|
||
bool isRial = unit is "ریال" or "ريال" || (unit.Length == 0 && hasRial);
|
||
if (isRial) v /= 10;
|
||
if (v > best) best = v;
|
||
}
|
||
|
||
// Sanity: a monthly figure of 200M+ تومان is implausible in Iran — if the ad never said
|
||
// «تومان», it was almost certainly ریال, so normalize.
|
||
if (best >= 200_000_000 && !hasToman) best /= 10;
|
||
return best > 0 ? best : null;
|
||
}
|
||
|
||
private static readonly Regex EmailRx = new(@"[A-Za-z0-9._%+\-]+@[A-Za-z0-9.\-]+\.[A-Za-z]{2,}", RegexOptions.Compiled);
|
||
private static readonly Regex UrlRx = new(@"https?://[^\s]+", RegexOptions.Compiled);
|
||
|
||
private static string ContactLabel(ContactType t) => ContactInfo.Label(t);
|
||
|
||
/// <summary>Pull every contact channel out of a post: phones, email, and socials (Instagram /
|
||
/// Telegram / Bale / WhatsApp / website) via URLs and Persian keyword cues.</summary>
|
||
private static List<ParsedContact> ExtractContacts(string raw)
|
||
{
|
||
var latin = ToLatinDigits(raw);
|
||
var list = new List<ParsedContact>();
|
||
void Add(ContactType t, string v)
|
||
{
|
||
v = v.Trim().Trim('.', '،', ',', ')', '(', ':', '«', '»', '"', '/').Trim();
|
||
if (v.Length < 2) return;
|
||
if (!list.Any(c => c.Type == t && string.Equals(c.Value, v, StringComparison.OrdinalIgnoreCase)))
|
||
list.Add(new ParsedContact(t, v));
|
||
}
|
||
|
||
foreach (Match m in EmailRx.Matches(latin)) Add(ContactType.Email, m.Value);
|
||
|
||
foreach (Match m in UrlRx.Matches(latin))
|
||
{
|
||
var u = m.Value.TrimEnd('.', '،', ')', '(', '"');
|
||
var low = u.ToLowerInvariant();
|
||
if (low.Contains("instagram.com") || low.Contains("instagr.am")) Add(ContactType.Instagram, UrlHandle(u));
|
||
else if (low.Contains("t.me") || low.Contains("telegram.me")) Add(ContactType.Telegram, UrlHandle(u));
|
||
else if (low.Contains("ble.ir") || low.Contains("bale.ai")) Add(ContactType.Bale, UrlHandle(u));
|
||
else if (low.Contains("wa.me") || low.Contains("whatsapp")) Add(ContactType.WhatsApp, UrlHandle(u));
|
||
else Add(ContactType.Website, u);
|
||
}
|
||
|
||
// Persian keyword → handle (latin handles only, so Persian words after the cue don't match).
|
||
void Keyed(ContactType t, params string[] kws)
|
||
{
|
||
foreach (var kw in kws)
|
||
foreach (Match m in Regex.Matches(latin, kw + @"\s*[:：]?\s*@?([A-Za-z0-9_.]{3,30})"))
|
||
Add(t, m.Groups[1].Value);
|
||
}
|
||
Keyed(ContactType.Instagram, "اینستاگرام", "اینستگرام", "اینستا", "پیج");
|
||
Keyed(ContactType.Telegram, "تلگرام");
|
||
Keyed(ContactType.WhatsApp, "واتساپ", "واتس اپ");
|
||
|
||
// phones — mobiles then landlines (multiple), boundary-guarded.
|
||
foreach (Match m in Regex.Matches(latin, @"(?<!\d)(?:\+?98|0)?9\d{9}(?!\d)"))
|
||
{
|
||
var d = Regex.Replace(m.Value, @"\D", "");
|
||
if (d.StartsWith("98")) d = "0" + d[2..];
|
||
if (d.Length == 10 && d[0] == '9') d = "0" + d;
|
||
Add(ContactType.Mobile, d);
|
||
}
|
||
// Landline area codes start 0[1-8] (021 Tehran, 026 Karaj, …) — never 09, which is a MOBILE.
|
||
// The old 0\d{2,3} matched 09xx numbers and mislabeled mobiles as «تلفن ثابت».
|
||
foreach (Match m in Regex.Matches(latin, @"(?<!\d)0[1-8]\d{1,2}[\s-]?\d{7,8}(?!\d)"))
|
||
Add(ContactType.Phone, Regex.Replace(m.Value, @"\D", ""));
|
||
|
||
return list.Take(8).ToList();
|
||
}
|
||
|
||
// Canonical tag → trigger words found in the post.
|
||
private static readonly (string Tag, string[] Needles)[] TagDict =
|
||
{
|
||
("mmt", new[] { "mmt", "ام ام تی", "امامتی" }),
|
||
("ICU", new[] { "icu", "آی سی یو", "آیسییو" }),
|
||
("CCU", new[] { "ccu", "سی سی یو", "سیسییو" }),
|
||
("NICU", new[] { "nicu", "ان آی سی یو", "نوزادان" }),
|
||
("BLS", new[] { "bls" }),
|
||
("ACLS", new[] { "acls" }),
|
||
("دیالیز", new[] { "دیالیز" }),
|
||
("اتاق عمل", new[] { "اتاق عمل", "اسکراب" }),
|
||
("بیهوشی", new[] { "بیهوشی" }),
|
||
("تریاژ", new[] { "تریاژ" }),
|
||
("تزریقات", new[] { "تزریقات", "تزریق" }),
|
||
("پانسمان", new[] { "پانسمان", "زخم" }),
|
||
("سونوگرافی", new[] { "سونوگرافی" }),
|
||
("رادیولوژی", new[] { "رادیولوژی" }),
|
||
("اورژانس", new[] { "اورژانس", "فوریت" }),
|
||
("مسئول فنی", new[] { "مسئول فنی" }),
|
||
("طرح", new[] { "طرح" }),
|
||
("سالمند", new[] { "سالمند" }),
|
||
("کودک", new[] { "کودک", "اطفال" }),
|
||
("همراه بیمار", new[] { "همراه بیمار" }),
|
||
("پروانهدار", new[] { "پروانه" }),
|
||
};
|
||
|
||
private static List<string> ExtractTags(string text)
|
||
{
|
||
var tags = new List<string>();
|
||
foreach (var (tag, needles) in TagDict)
|
||
if (ContainsAny(text, needles)) tags.Add(tag);
|
||
return tags;
|
||
}
|
||
|
||
private static string UrlHandle(string url)
|
||
{
|
||
var u = url.Split('?')[0].TrimEnd('/');
|
||
var seg = u.Contains('/') ? u[(u.LastIndexOf('/') + 1)..] : u;
|
||
return string.IsNullOrWhiteSpace(seg) ? url : seg;
|
||
}
|
||
|
||
private static string Normalize(string s) => s
|
||
.Replace('ي', 'ی').Replace('ك', 'ک').Replace('', ' ').Trim();
|
||
|
||
private static bool ContainsAny(string text, params string[] needles)
|
||
=> needles.Any(n => text.Contains(n));
|
||
|
||
private static string ToLatinDigits(string s)
|
||
{
|
||
var chars = s.ToCharArray();
|
||
for (var i = 0; i < chars.Length; i++)
|
||
{
|
||
if (chars[i] >= '۰' && chars[i] <= '۹') chars[i] = (char)('0' + (chars[i] - '۰'));
|
||
else if (chars[i] >= '٠' && chars[i] <= '٩') chars[i] = (char)('0' + (chars[i] - '٠'));
|
||
}
|
||
return new string(chars);
|
||
}
|
||
}
|