Harden 8-K item-code classifier against dirty input strings

The _classify_event_type mapping (2.02 → earnings_release, 7.01 →
guidance_update, 1.01 → material_contract, 1.03 → other_material_event,
8.01 → other_material_event, 5.02 → management_change) was already in
place but used naive 'in items' string matching. Upstream extractors
sometimes deliver items as 'Item 2.02' or '2.02 - Results of Operations'
(full-description form), which silently slipped through to event_type
'unknown' and were rejected by all 12 v7.356 PEAD engines.

A reparse using the patched classifier touched 9,779 historical 'unknown'
rows; only 16 actually flipped (the rest are genuinely off-vocab 8-Ks
like 9.01-only, 3.01, 5.07). The fix is therefore small in retroactive
impact, but defends against future ingestion drift.

Changes:
 - libs/parser/rule_parser.py: rewrote _classify_event_type with
   _normalize_item_codes (regex \\b(\\d+\\.\\d+)\\b token extractor) and
   tuple-of-pairs _ITEM_TO_EVENT_TYPE mapping. Earnings_release wins
   priority over management_change when 2.02 + 5.02 co-occur, consistent
   with the strategy's vocabulary intent.
 - tests/unit/test_rule_parser.py: 7 new regression tests covering
   AMD/MNST-style 2.02+9.01, dirty 'Item 2.02' / '2.02 - Results...'
   forms, and negative cases (9.01-only, 2.03, 3.01 remain unknown).

Note: a follow-up vocabulary normalizer is still needed for the
Oracle-fallback path in apps/pipeline/event_parser/main.py:140, which
writes raw oracle_event.event_type values like 'earnings_result',
'shareholder_vote', 'regulation_fd' that don't match the strategy
vocabulary. Flagged for separate ticket.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

libs/parser/rule_parser.py

@@ -129,19 +129,54 @@ def _extract_item_numbers(text: str) -> list[str]:
     return list(dict.fromkeys(_ITEM_RE.findall(text)))
 
 
+# Pattern used to harvest a clean N.NN item code from an arbitrarily formatted
+# entry such as "2.02", "Item 2.02", "2.02 - Results of Operations and ...",
+# or "Item 2.02 Results of Operations and Financial Condition".
+_ITEM_NUMBER_TOKEN_RE = re.compile(r"\b(\d+\.\d+)\b")
+
+
+def _normalize_item_codes(items: list[str]) -> list[str]:
+    """Normalize raw item entries into a list of clean N.NN codes.
+
+    Tolerates upstream variation:
+      - "2.02"                                     -> "2.02"
+      - "Item 2.02"                                -> "2.02"
+      - "2.02 - Results of Operations"             -> "2.02"
+      - "Item 2.02 Results of Operations"          -> "2.02"
+      - "Results of Operations and Financial..."   -> "" (dropped)
+    Order-preserving, deduplicated.
+    """
+    normalized: list[str] = []
+    for raw in items or []:
+        if not isinstance(raw, str):
+            continue
+        m = _ITEM_NUMBER_TOKEN_RE.search(raw)
+        if m:
+            code = m.group(1)
+            if code not in normalized:
+                normalized.append(code)
+    return normalized
+
+
+# 8-K item code -> ParserEventOutput.event_type vocabulary.
+# Earnings (2.02) takes precedence over downstream items like 9.01 because a
+# 2.02 release is conventionally co-filed with 9.01 (Financial Statements and
+# Exhibits) — the strategy must see those as earnings_release, not "unknown".
+_ITEM_TO_EVENT_TYPE: tuple[tuple[str, str], ...] = (
+    ("2.02", "earnings_release"),
+    ("7.01", "guidance_update"),
+    ("5.02", "management_change"),
+    ("1.01", "material_contract"),
+    ("1.03", "other_material_event"),
+    ("8.01", "other_material_event"),
+)
+
+
 def _classify_event_type(items: list[str]) -> str:
-    if "2.02" in items:
+    codes = _normalize_item_codes(items)
-        return "earnings_release"
+    for code, event_type in _ITEM_TO_EVENT_TYPE:
-    if "7.01" in items:
+        if code in codes:
-        return "guidance_update"
+            return event_type
-    if "1.01" in items:
-        return "material_contract"
-    if "8.01" in items:
-        return "other_material_event"
-    if "5.02" in items:
-        return "management_change"
-    if "1.03" in items:
-        return "other_material_event"
     return "unknown"
 
 

tests/unit/test_rule_parser.py

@@ -165,3 +165,124 @@ def test_positive_backlog_and_customer_signals_outweigh_single_financing_flag():
     out = p.parse("DOC::test", "8-K", text, {"filing_date": "2026-03-05"})
 
     assert out.event_direction == "bullish"
+
+
+# ---------------------------------------------------------------------------
+# Item -> event_type classification (regression tests for unknown-class bug
+# where AMD/MNST 2.02 + 9.01 earnings filings were tagged "unknown").
+# ---------------------------------------------------------------------------
+
+
+def test_classify_event_type_amd_style_earnings_release_with_9_01():
+    """AMD/MNST file 2.02 alongside 9.01; must classify as earnings_release."""
+    from libs.parser.rule_parser import _classify_event_type
+
+    # Order in either direction — earnings still wins.
+    assert _classify_event_type(["2.02", "9.01"]) == "earnings_release"
+    assert _classify_event_type(["9.01", "2.02"]) == "earnings_release"
+
+
+def test_classify_event_type_tolerates_dirty_item_strings():
+    """Upstream may store entries like 'Item 2.02' or '2.02 - Results...'."""
+    from libs.parser.rule_parser import _classify_event_type
+
+    assert _classify_event_type(["Item 2.02", "Item 9.01"]) == "earnings_release"
+    assert (
+        _classify_event_type(
+            ["Item 2.02 Results of Operations and Financial Condition"]
+        )
+        == "earnings_release"
+    )
+    assert _classify_event_type(["2.02 - Results of Operations"]) == "earnings_release"
+
+
+def test_classify_event_type_known_mappings_each_item():
+    from libs.parser.rule_parser import _classify_event_type
+
+    assert _classify_event_type(["7.01"]) == "guidance_update"
+    assert _classify_event_type(["1.01"]) == "material_contract"
+    assert _classify_event_type(["8.01"]) == "other_material_event"
+    assert _classify_event_type(["1.03"]) == "other_material_event"
+    # 5.02 (departure of officers) -> management_change, even when co-filed
+    # with 9.01 financial-statements-and-exhibits.
+    assert _classify_event_type(["5.02", "9.01"]) == "management_change"
+
+
+def test_classify_event_type_negative_cases_remain_unknown():
+    """Items that aren't in the strategy vocabulary must remain 'unknown'.
+
+    This guards against accidental over-eager relabeling.
+    """
+    from libs.parser.rule_parser import _classify_event_type
+
+    # 9.01 alone (financial statements & exhibits) is not its own event type
+    assert _classify_event_type(["9.01"]) == "unknown"
+    # 2.03 (financial obligation), 3.01 (delisting), 5.07 (shareholder vote)
+    assert _classify_event_type(["2.03"]) == "unknown"
+    assert _classify_event_type(["3.01"]) == "unknown"
+    assert _classify_event_type(["5.07"]) == "unknown"
+    # No items at all
+    assert _classify_event_type([]) == "unknown"
+    # Description-only with no extractable code
+    assert (
+        _classify_event_type(["Results of Operations and Financial Condition"])
+        == "unknown"
+    )
+
+
+def test_parse_amd_style_earnings_with_only_9_01_in_item_numbers_uses_text_fallback():
+    """If the SGML loss left only 9.01 in item_numbers but the body says
+    'AMD Reports Q3 Financial Results', the body regex must catch it."""
+    from libs.parser.rule_parser import RuleBasedParser
+    from libs.parser.text_normalizer import normalize_text
+
+    p = RuleBasedParser()
+    html = """
+    <html><body>
+    <div>NEWS RELEASE</div>
+    <div>AMD Reports Third Quarter 2025 Financial Results</div>
+    <div>Revenue of $7.7 billion grew year over year.</div>
+    </body></html>
+    """
+    out = p.parse(
+        "DOC::test",
+        "8-K",
+        normalize_text(html, is_html=True),
+        {"filing_date": "2025-10-29", "item_numbers": ["9.01"]},
+    )
+    assert out.event_type == "earnings_release"
+
+
+def test_parse_mnst_style_earnings_dirty_item_format_classifies_correctly():
+    """MNST-style: 'Item 2.02 Results of Operations and Financial Condition'
+    plus 'Item 9.01 ...' as raw entries must classify earnings_release."""
+    from libs.parser.rule_parser import RuleBasedParser
+
+    p = RuleBasedParser()
+    text = (
+        "Monster Beverage Corporation Reports 2025 Third Quarter Financial Results\n"
+        "Net sales of $1.97 billion."
+    )
+    out = p.parse(
+        "DOC::test",
+        "8-K",
+        text,
+        {
+            "filing_date": "2025-11-06",
+            "item_numbers": [
+                "Item 2.02 Results of Operations and Financial Condition",
+                "Item 9.01 Financial Statements and Exhibits",
+            ],
+        },
+    )
+    assert out.event_type == "earnings_release"
+
+
+def test_normalize_item_codes_drops_non_string_and_dedupes():
+    from libs.parser.rule_parser import _normalize_item_codes
+
+    # Order-preserving + dedup + non-string ignored
+    assert _normalize_item_codes(["Item 2.02", "2.02", "9.01", None, 7]) == [
+        "2.02",
+        "9.01",
+    ]