Serik B. Malikov
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Serik B. Malikov has a total of 3 co-authors. Here are the top co-authors by number of shared publications:
- Aiman Bekmuratovna Omarova
Narxoz University
1 shared publication - Ainash Seitzhan
Narxoz University, Almaty 050035, Kazakhstan
1 shared publication - Sholpan Baltabekovna Malikova
Narxoz University, Almaty 050035, Kazakhstan
1 shared publication
Latest Publications
- Open Access
What Explicit Rules and Cross-Lingual Similarity Can and Cannot Check in Kazakh–Russian Criminal Legislation: A Benchmark with a Blinded Expert Review
Kazakhstan enacts its laws in Kazakh and Russian, both authentic, yet the two texts are not compared when the state’s reference database of legal acts is formed. On the Criminal Code, this benchmark tests cross-lingual similarity and explicit rules against synthetic one-sided edits, differences between editions and a blinded review of 192 units by two legal experts and an arbiter. The best models matched 98.6–100% of 487 articles to their counterparts, which is not detection. A global similarity threshold detected edits weakly (area under the receiver operating characteristic curve 0.49–0.67; at most 0.81 after length calibration, still missing over half of the changed references at a 5% false-positive rate), whereas comparing a unit with its previous text registered most edits. No similarity score ranked expert-identified discrepancies above chance after adjustment for multiple comparisons. Prototype rules on structure, numbers, references and units flagged 16 articles, 10 with a discrepancy, six at the flagged element. Of 150 random unflagged units, eight (5.3%) held a discrepancy, mostly in terms or the scope of qualifiers, which no rule reads; the flags thus pointed to an estimated 7% (4–13%) of the units with discrepancies (at most 14% under other expert references). On a second code, only three of 52 and seven of 34 flags of the two quantity readers marked real differences. The rules can guide expert review on the code they were built for but do not transfer without adaptation and cannot replace it.
Open AccessWhat Explicit Rules and Cross-Lingual Similarity Can and Cannot Check in Kazakh–Russian Criminal Legislation: A Benchmark with a Blinded Expert Review
Kazakhstan enacts its laws in Kazakh and Russian, both authentic, yet the two texts are not compared when the state’s reference database of legal acts is formed. On the Criminal Code, this benchmark tests cross-lingual similarity and explicit rules against synthetic one-sided edits, differences between editions and a blinded review of 192 units by two legal experts and an arbiter. The best models matched 98.6–100% of 487 articles to their counterparts, which is not detection. A global similarity threshold detected edits weakly (area under the receiver operating characteristic curve 0.49–0.67; at most 0.81 after length calibration, still missing over half of the changed references at a 5% false-positive rate), whereas comparing a unit with its previous text registered most edits. No similarity score ranked expert-identified discrepancies above chance after adjustment for multiple comparisons. Prototype rules on structure, numbers, references and units flagged 16 articles, 10 with a discrepancy, six at the flagged element. Of 150 random unflagged units, eight (5.3%) held a discrepancy, mostly in terms or the scope of qualifiers, which no rule reads; the flags thus pointed to an estimated 7% (4–13%) of the units with discrepancies (at most 14% under other expert references). On a second code, only three of 52 and seven of 34 flags of the two quantity readers marked real differences. The rules can guide expert review on the code they were built for but do not transfer without adaptation and cannot replace it.