Improve classifier fallback for flat CV parsing

2026-04-01 11:00:53 +02:00
parent 517c42250d
commit b283f8b9d2
2 changed files with 274 additions and 29 deletions
@@ -617,9 +617,47 @@ public sealed class ProfileCvControllerTests
        Assert.Contains("Atlas Systems", matchedJob!.Company ?? string.Empty, StringComparison.Ordinal);
        Assert.Contains("Python", structured.Skills);
        Assert.Contains("SQL", structured.Skills);
+        Assert.Equal("classifier", structured.Metadata.Fields["jobs[0].title"].Method);
+        Assert.Equal("block-1", structured.Metadata.Fields["jobs[0].title"].SourceBlockId);
        classifier.Verify(x => x.ClassifyBlockAsync(It.IsAny<string>(), It.IsAny<CancellationToken>()), Times.AtLeastOnce());
    }

+    [Fact]
+    public async Task Parse_uses_classifier_fallback_for_education_blocks_without_real_sections()
+    {
+        var source = "BSc Computer Science\nUniversity of Oslo\nOslo\n2016 - 2019\nGraduated with focus on distributed systems.";
+        var user = new ApplicationUser { Id = "user-1", ProfileCvText = source };
+        var userManager = CreateUserManager();
+        userManager.Setup(x => x.GetUserAsync(It.IsAny<ClaimsPrincipal>())).ReturnsAsync(user);
+        userManager.Setup(x => x.UpdateAsync(user)).ReturnsAsync(IdentityResult.Success);
+        var aiService = new Mock<ISummarizerService>();
+        aiService
+            .Setup(x => x.SummarizeSectionAsync(It.Is<string>(instruction => instruction.Contains("Extract this CV into structured JSON", StringComparison.Ordinal)), source, 3200, 900))
+            .ReturnsAsync("not-json");
+
+        var classifier = new Mock<ICvAiClassifier>();
+        classifier
+            .Setup(x => x.ClassifyBlockAsync(It.IsAny<string>(), It.IsAny<CancellationToken>()))
+            .ReturnsAsync(new CvBlockClassificationResult("Education", 0.87, "education block", "BSc Computer Science", "University of Oslo", "Oslo", "2016", "2019", new List<string> { "Graduated with focus on distributed systems." }));
+
+        await using var db = CreateDb();
+        var paths = CreatePaths();
+        var controller = CreateController(userManager.Object, aiService.Object, db, paths, classifier.Object);
+
+        var result = await controller.Parse(new ProfileCvController.ParseCvRequest(source));
+
+        var ok = Assert.IsType<OkObjectResult>(result.Result);
+        var json = JsonSerializer.Serialize(ok.Value);
+        Assert.Contains("BSc Computer Science", json);
+        Assert.Contains("University of Oslo", json);
+
+        var structured = StructuredCvProfileJson.Deserialize(user.ProfileCvStructureJson);
+        Assert.Single(structured.Education);
+        Assert.Equal("BSc Computer Science", structured.Education[0].Qualification);
+        Assert.Equal("University of Oslo", structured.Education[0].Institution);
+        Assert.Equal("classifier", structured.Metadata.Fields["education[0].qualification"].Method);
+    }
+
    [Fact]
    public async Task Parse_keeps_general_fallback_when_classifier_returns_nothing()
    {
@@ -78,6 +78,7 @@ public sealed class ProfileCvController : ControllerBase
    public sealed record ParseCvRequest(string? Text);

    private sealed record ExtractionPipelineResult(string RawText, string NormalizedText, StructuredCvProfile StructuredCv);
+    private sealed record ClassifiedCvBlock(int Index, string OriginalBlock, string SectionName, string Content, CvBlockClassificationResult? Classification);
    public sealed record CvExtractionRunListItem(
        int Id,
        string Trigger,
@@ -340,7 +341,30 @@ public sealed class ProfileCvController : ControllerBase
    private async Task<StructuredCvProfile> BuildStructuredCvAsync(string text, CancellationToken cancellationToken)
    {
        var parseSource = NormalizeTextForStructuredParsing(text);
-        var fallbackSections = await BuildFallbackSectionsAsync(parseSource, cancellationToken);
+        var parsedSections = ParseSections(parseSource)
+            .Select(section => new StructuredCvSection
+            {
+                Name = section.Name,
+                Content = section.Content,
+                WordCount = CountWords(section.Content),
+            })
+            .ToList();
+        var hasRealSections = parsedSections.Any(section => !string.Equals(section.Name, "General", StringComparison.OrdinalIgnoreCase));
+
+        List<ClassifiedCvBlock> classifiedBlocks = new();
+        List<StructuredCvSection> fallbackSections = parsedSections;
+        StructuredCvProfile? classifierFallback = null;
+
+        if (!hasRealSections)
+        {
+            classifiedBlocks = await ClassifyBlocksAsync(parseSource, cancellationToken);
+            var hasMeaningfulClassifierStructure = classifiedBlocks.Any(block => !string.Equals(block.SectionName, "General", StringComparison.OrdinalIgnoreCase));
+            if (hasMeaningfulClassifierStructure)
+            {
+                fallbackSections = BuildSectionsFromClassifiedBlocks(classifiedBlocks);
+                classifierFallback = BuildStructuredCvFromClassifiedBlocks(classifiedBlocks);
+            }
+        }

        var sectionFallback = StructuredCvProfileJson.FromSections(fallbackSections);
        AnnotateStructuredCv(sectionFallback, "repair", 0.56);
@@ -348,6 +372,10 @@ public sealed class ProfileCvController : ControllerBase
        AnnotateStructuredCv(heuristicFallback, "deterministic", 0.68);
        heuristicFallback.Sections = new List<StructuredCvSection>();
        var fallback = StructuredCvProfileJson.Merge(heuristicFallback, sectionFallback);
+        if (classifierFallback is not null)
+        {
+            fallback = StructuredCvProfileJson.Merge(classifierFallback, fallback);
+        }
        fallback.Contact.FullName ??= GuessFullName(text) ?? GuessFullNameFromEmail(fallback.Contact.Email);
        var extracted = await TryExtractStructuredCvAsync(parseSource, cancellationToken);
        var merged = StructuredCvProfileJson.Merge(extracted, fallback);
@@ -874,36 +902,202 @@ public sealed class ProfileCvController : ControllerBase
            .ToList();
    }

-    private async Task<List<StructuredCvSection>> BuildFallbackSectionsAsync(string parseSource, CancellationToken cancellationToken)
+    private static List<StructuredCvSection> BuildSectionsFromClassifiedBlocks(List<ClassifiedCvBlock> classifiedBlocks)
    {
-        var parsed = ParseSections(parseSource)
-            .Select(section => new StructuredCvSection
+        var sectionBuckets = new List<StructuredCvSection>();
+        foreach (var block in classifiedBlocks)
        {
-                Name = section.Name,
-                Content = section.Content,
-                WordCount = CountWords(section.Content),
-            })
-            .ToList();
-
-        var hasRealSections = parsed.Any(section => !string.Equals(section.Name, "General", StringComparison.OrdinalIgnoreCase));
-        if (hasRealSections) return parsed;
-
-        var aiSections = await ClassifyBlocksIntoSectionsAsync(parseSource, cancellationToken);
-        return aiSections.Count > 0 ? aiSections : parsed;
+            var existing = sectionBuckets.FirstOrDefault(section => section.Name == block.SectionName);
+            if (existing is null)
+            {
+                sectionBuckets.Add(new StructuredCvSection { Name = block.SectionName, Content = block.Content, WordCount = CountWords(block.Content) });
+            }
+            else
+            {
+                existing.Content = $"{existing.Content}\n\n{block.Content}".Trim();
+                existing.WordCount = CountWords(existing.Content);
+            }
        }

-    private async Task<List<StructuredCvSection>> ClassifyBlocksIntoSectionsAsync(string parseSource, CancellationToken cancellationToken)
+        return sectionBuckets.Where(section => !string.IsNullOrWhiteSpace(section.Content)).ToList();
+    }
+
+    private static StructuredCvProfile BuildStructuredCvFromClassifiedBlocks(List<ClassifiedCvBlock> classifiedBlocks)
+    {
+        var profile = new StructuredCvProfile();
+        var now = DateTimeOffset.UtcNow;
+        var summary = new List<string>();
+        var skills = new HashSet<string>(StringComparer.OrdinalIgnoreCase);
+
+        foreach (var block in classifiedBlocks)
+        {
+            switch (block.SectionName)
+            {
+                case "Professional Summary":
+                    foreach (var item in SplitClassifierContent(block.Content, 5))
+                    {
+                        summary.Add(item);
+                    }
+                    ApplyClassifierFieldMetadata(profile, "summary", summary.FirstOrDefault(), block, now);
+                    break;
+                case "Skills":
+                    foreach (var item in SplitClassifierSkills(block.Content))
+                    {
+                        skills.Add(item);
+                    }
+                    ApplyClassifierFieldMetadata(profile, "skills", skills.FirstOrDefault(), block, now);
+                    break;
+                case "Work Experience":
+                    var job = BuildJobFromClassifiedBlock(block);
+                    if (job is not null)
+                    {
+                        var index = profile.Jobs.Count;
+                        profile.Jobs.Add(job);
+                        ApplyClassifierFieldMetadata(profile, $"jobs[{index}].title", job.Title, block, now);
+                        ApplyClassifierFieldMetadata(profile, $"jobs[{index}].company", job.Company, block, now);
+                        ApplyClassifierFieldMetadata(profile, $"jobs[{index}].location", job.Location, block, now);
+                    }
+                    break;
+                case "Education":
+                    var education = BuildEducationFromClassifiedBlock(block);
+                    if (education is not null)
+                    {
+                        var index = profile.Education.Count;
+                        profile.Education.Add(education);
+                        ApplyClassifierFieldMetadata(profile, $"education[{index}].qualification", education.Qualification, block, now);
+                        ApplyClassifierFieldMetadata(profile, $"education[{index}].institution", education.Institution, block, now);
+                    }
+                    break;
+                default:
+                    if (!string.IsNullOrWhiteSpace(block.Content))
+                    {
+                        profile.OtherSections.Add(new StructuredCvOtherSection
+                        {
+                            Title = block.SectionName,
+                            Items = SplitClassifierContent(block.Content, 6)
+                        });
+                    }
+                    break;
+            }
+        }
+
+        profile.Summary = summary.Distinct(StringComparer.OrdinalIgnoreCase).ToList();
+        profile.Skills = skills.ToList();
+        profile.Sections = BuildSectionsFromClassifiedBlocks(classifiedBlocks);
+
+        var averageConfidence = classifiedBlocks
+            .Select(block => block.Classification?.Confidence)
+            .Where(value => value.HasValue)
+            .Select(value => value!.Value)
+            .DefaultIfEmpty(0.74)
+            .Average();
+        AnnotateStructuredCv(profile, "classifier", averageConfidence);
+        return StructuredCvProfileJson.Normalize(profile);
+    }
+
+    private static StructuredCvJob? BuildJobFromClassifiedBlock(ClassifiedCvBlock block)
+    {
+        var classification = block.Classification;
+        if (classification is null) return null;
+
+        var bullets = classification.Bullets is { Count: > 0 }
+            ? classification.Bullets.Where(bullet => !string.IsNullOrWhiteSpace(bullet)).Select(bullet => bullet.Trim()).ToList()
+            : SplitClassifierContent(block.OriginalBlock, 6);
+
+        var job = new StructuredCvJob
+        {
+            Title = NullIfWhitespace(classification.Title),
+            Company = NullIfWhitespace(classification.Company),
+            Location = NullIfWhitespace(classification.Location),
+            Start = NullIfWhitespace(classification.Start),
+            End = NullIfWhitespace(classification.End),
+            IsCurrent = string.Equals(classification.End, "Present", StringComparison.OrdinalIgnoreCase) || string.Equals(classification.End, "Current", StringComparison.OrdinalIgnoreCase),
+            Bullets = bullets,
+            Skills = SplitClassifierSkills(block.OriginalBlock)
+        };
+
+        return StructuredCvProfileJson.Normalize(new StructuredCvProfile { Jobs = new List<StructuredCvJob> { job } }).Jobs.FirstOrDefault();
+    }
+
+    private static StructuredCvEducation? BuildEducationFromClassifiedBlock(ClassifiedCvBlock block)
+    {
+        var classification = block.Classification;
+        if (classification is null) return null;
+
+        var education = new StructuredCvEducation
+        {
+            Qualification = NullIfWhitespace(classification.Title),
+            Institution = NullIfWhitespace(classification.Company),
+            Location = NullIfWhitespace(classification.Location),
+            Start = NullIfWhitespace(classification.Start),
+            End = NullIfWhitespace(classification.End),
+            Details = classification.Bullets is { Count: > 0 }
+                ? classification.Bullets.Where(bullet => !string.IsNullOrWhiteSpace(bullet)).Select(bullet => bullet.Trim()).ToList()
+                : SplitClassifierContent(block.OriginalBlock, 5)
+        };
+
+        return StructuredCvProfileJson.Normalize(new StructuredCvProfile { Education = new List<StructuredCvEducation> { education } }).Education.FirstOrDefault();
+    }
+
+    private static List<string> SplitClassifierContent(string content, int limit)
+    {
+        return content
+            .Replace("\r\n", "\n")
+            .Split(new[] { '\n', '•' }, StringSplitOptions.RemoveEmptyEntries | StringSplitOptions.TrimEntries)
+            .SelectMany(line => line.Contains(". ", StringComparison.Ordinal)
+                ? Regex.Split(line, @"(?<=[.!?])\s+")
+                : new[] { line })
+            .Select(item => item.Trim().TrimStart('-', '•', '*', '+', ' '))
+            .Where(item => item.Length > 2)
+            .Take(limit)
+            .Distinct(StringComparer.OrdinalIgnoreCase)
+            .ToList();
+    }
+
+    private static List<string> SplitClassifierSkills(string content)
+    {
+        return content
+            .Replace("\r\n", "\n")
+            .Split(new[] { '\n', ',', ';', '•' }, StringSplitOptions.RemoveEmptyEntries | StringSplitOptions.TrimEntries)
+            .Select(item => item.Trim().TrimStart('-', '•', '*', '+', ' '))
+            .Where(item => item.Length > 1 && item.Length <= 48 && !LooksLikeDateLikeValue(item) && !item.Contains('@'))
+            .Distinct(StringComparer.OrdinalIgnoreCase)
+            .ToList();
+    }
+
+    private static bool LooksLikeDateLikeValue(string value)
+    {
+        return Regex.IsMatch(value, @"^(?:\d{4}|(?:Jan|January|Feb|February|Mar|March|Apr|April|May|Jun|June|Jul|July|Aug|August|Sep|Sept|September|Oct|October|Nov|November|Dec|December)\s+\d{4}|Present|Current)(?:\s*[-–]\s*(?:\d{4}|(?:Jan|January|Feb|February|Mar|March|Apr|April|May|Jun|June|Jul|July|Aug|August|Sep|Sept|September|Oct|October|Nov|November|Dec|December)\s+\d{4}|Present|Current))?$", RegexOptions.IgnoreCase);
+    }
+
+    private static void ApplyClassifierFieldMetadata(StructuredCvProfile profile, string key, string? value, ClassifiedCvBlock block, DateTimeOffset now)
+    {
+        if (string.IsNullOrWhiteSpace(value)) return;
+
+        profile.Metadata.Fields[key] = new StructuredCvFieldMetadata
+        {
+            Confidence = block.Classification?.Confidence ?? 0.74,
+            Method = "classifier",
+            SourceSnippet = block.OriginalBlock.Length > 180 ? block.OriginalBlock[..180] : block.OriginalBlock,
+            SourceBlockId = $"block-{block.Index}",
+            ReviewState = string.Equals(block.SectionName, "General", StringComparison.OrdinalIgnoreCase) ? "needs-review" : "suggested",
+            LastUpdatedAtUtc = now,
+        };
+    }
+
+    private async Task<List<ClassifiedCvBlock>> ClassifyBlocksAsync(string parseSource, CancellationToken cancellationToken)
    {
        var blocks = Regex.Split(parseSource.Replace("\r\n", "\n"), @"\n\s*\n")
            .Select(block => block.Trim())
            .Where(block => block.Length >= 24)
            .ToList();

-        if (blocks.Count == 0) return new List<StructuredCvSection>();
+        if (blocks.Count == 0) return new List<ClassifiedCvBlock>();

-        var sectionBuckets = new List<StructuredCvSection>();
-        foreach (var block in blocks)
+        var results = new List<ClassifiedCvBlock>();
+        for (var index = 0; index < blocks.Count; index++)
        {
+            var block = blocks[index];
            var classification = await _cvAiClassifier.ClassifyBlockAsync(block, cancellationToken);
            var sectionName = classification?.Section;
            if (!string.IsNullOrWhiteSpace(sectionName) && SectionAliases.TryGetValue(sectionName, out var canonical))
@@ -931,20 +1125,33 @@ public sealed class ProfileCvController : ControllerBase
                }
                if (lines.Count > 0) content = string.Join("\n", lines);
            }
-
-            var existing = sectionBuckets.FirstOrDefault(section => section.Name == sectionName);
-            if (existing is null)
+            else if (string.Equals(sectionName, "Education", StringComparison.OrdinalIgnoreCase) && classification is not null)
            {
-                sectionBuckets.Add(new StructuredCvSection { Name = sectionName, Content = content, WordCount = CountWords(content) });
-            }
-            else
+                var lines = new List<string>();
+                if (!string.IsNullOrWhiteSpace(classification.Title)) lines.Add($"### {classification.Title.Trim()}");
+                var dateRange = FormatDateRangeForSection(classification.Start, classification.End, false);
+                var meta = string.Join(" | ", new[] { classification.Company, classification.Location, dateRange }.Where(value => !string.IsNullOrWhiteSpace(value)));
+                if (!string.IsNullOrWhiteSpace(meta)) lines.Add(meta);
+                if (classification.Bullets is not null)
                {
-                existing.Content = $"{existing.Content}\n\n{content}".Trim();
-                existing.WordCount = CountWords(existing.Content);
+                    lines.AddRange(classification.Bullets.Where(bullet => !string.IsNullOrWhiteSpace(bullet)).Select(bullet => $"- {bullet.Trim()}"));
                }
+                if (lines.Count > 0) content = string.Join("\n", lines);
+            }
+            else if (string.Equals(sectionName, "Skills", StringComparison.OrdinalIgnoreCase))
+            {
+                var items = SplitClassifierSkills(block);
+                if (items.Count > 0) content = string.Join("\n", items);
+            }
+            else if (string.Equals(sectionName, "Professional Summary", StringComparison.OrdinalIgnoreCase) && classification?.Bullets is { Count: > 0 })
+            {
+                content = string.Join("\n", classification.Bullets.Where(bullet => !string.IsNullOrWhiteSpace(bullet)).Select(bullet => $"- {bullet.Trim()}"));
            }

-        return sectionBuckets.Where(section => !string.IsNullOrWhiteSpace(section.Content)).ToList();
+            results.Add(new ClassifiedCvBlock(index + 1, block, sectionName, content, classification));
+        }
+
+        return results;
    }

    private static string? FormatDateRangeForSection(string? start, string? end, bool isCurrent)