{"messages": [{"role": "system", "content": "You are assisting with linguistic annotation for historical and under-resourced languages.\n\nYou must follow the annotation guidelines exactly.\n\nYour task is not to translate, paraphrase, normalise, or improve the input text. Your task is to annotate the tokens provided by the user.\n\nAlways preserve the input tokenisation. Do not add, remove, split, merge, reorder, transliterate, or translate tokens.\n\nWhen uncertain, use the authorised uncertainty mechanism rather than inventing an answer.\n\nReturn only the requested structured output. Do not include explanations outside the requested JSON.\n"}, {"role": "user", "content": "Annotate the following pre-tokenised sentence token by token.\n\nYou must follow the same annotation conventions as in the validated examples.\n\nLanguage: Ancient Greek\nScript: (unspecified)\nDomain: (unspecified)\nSentence ID: s2\n\nValidated examples:\n(none)\n\nNow annotate this sentence.\n\nOriginal text:\nμετοικήσαντες πόλιν, ἕως ἂν τὸ τοῦ χρόνου μῆκος ταύτην αὐτοῖς\n\nInput tokens:\n[\"μετοικήσαντες\", \"πόλιν,\", \"ἕως\", \"ἂν\", \"τὸ\", \"τοῦ\", \"χρόνου\", \"μῆκος\", \"ταύτην\", \"αὐτοῖς\"]\n\nTask:\nFor each input token, provide:\n1. the original surface form;\n2. a lemma;\n3. a UPOS tag;\n4. selected morphological features;\n5. a confidence value;\n6. an optional short local comment.\n\nAuthorised UPOS inventory:\n- `lemma`: free string (nullable)\n- `pos` ∈ {A, I+Adv, I+AdvPr, I+Conj, I+Intj, I+Neg, I+Part, I+Prep, LF, N+Ant, N+Com, N+Epi, N+Let, N+Pat, N+Prop, N+Top, NUM+Car, NUM+Ord, NUMA+Car, NUMA+Ord, PRO+Dem, PRO+Ind, PRO+Int, PRO+Per1p, PRO+Per1s, PRO+Per2p, PRO+Per2s, PRO+Pos1p, PRO+Pos1s, PRO+Pos2p, PRO+Pos2s, PRO+Rec, PRO+Ref, PRO+Rel, V} — or any combination of these tags joined by `@` (e.g. `TAG1@TAG2`).\n- `confidence` ∈ {low, medium, high}\n- `comment`: free string (nullable)\n\nAuthorised morphological features:\n(no morphological subfields)\n\nRules:\n- Return only valid JSON.\n- Do not translate the sentence.\n- Do not transliterate the tokens.\n- Do not normalise the surface forms.\n- Do not add, remove, split, merge, or reorder tokens.\n- The output must contain exactly one annotation per input token.\n- The value of `surface` must exactly match the corresponding input token.\n- Use only the authorised UPOS tags.\n- Use only the authorised feature names and values.\n- Follow the lemma conventions shown in the validated examples.\n- Follow the treatment of uncertain or ambiguous forms shown in the validated examples.\n- If uncertain, set `confidence` to `\"low\"` or `\"medium\"` and add a short local comment.\n- Keep `comment` short and local. Do not provide a full translation.\n\nExpected output schema:\n{\n  \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n  \"title\": \"HYE POS + lemma (bespoke tagset) annotation\",\n  \"type\": \"object\",\n  \"required\": [\n    \"sentence_id\",\n    \"language\",\n    \"tokens\"\n  ],\n  \"additionalProperties\": false,\n  \"properties\": {\n    \"sentence_id\": {\n      \"type\": \"string\"\n    },\n    \"language\": {\n      \"type\": \"string\"\n    },\n    \"tokens\": {\n      \"type\": \"array\",\n      \"items\": {\n        \"type\": \"object\",\n        \"required\": [\n          \"surface\",\n          \"lemma\",\n          \"pos\",\n          \"confidence\",\n          \"comment\"\n        ],\n        \"additionalProperties\": false,\n        \"properties\": {\n          \"surface\": {\n            \"type\": \"string\"\n          },\n          \"lemma\": {\n            \"type\": [\n              \"string\",\n              \"null\"\n            ]\n          },\n          \"pos\": {\n            \"type\": \"string\"\n          },\n          \"confidence\": {\n            \"type\": \"string\",\n            \"enum\": [\n              \"low\",\n              \"medium\",\n              \"high\"\n            ]\n          },\n          \"comment\": {\n            \"type\": [\n              \"string\",\n              \"null\"\n            ]\n          }\n        }\n      }\n    }\n  }\n}\n"}, {"role": "assistant", "content": "{\"sentence_id\": \"s1\", \"language\": \"Ancient Greek\", \"tokens\": [{\"surface\": \"ΙΩΑΝΝΟΥ\"}, {\"surface\": \"ΑΝΑΓΝΩΣΤΟΥ\"}, {\"surface\": \"Διήγησις\"}, {\"surface\": \"περὶ\"}, {\"surface\": \"τῆς\"}, {\"surface\": \"τελευταίας\"}, {\"surface\": \"ἁλώσεως\"}, {\"surface\": \"τῆς\"}, {\"surface\": \"Θεσσαλονίκης\"}, {\"surface\": \"συντεθεῖσα\"}]}"}]}
{"messages": [{"role": "system", "content": "You are assisting with linguistic annotation for historical and under-resourced languages.\n\nYou must follow the annotation guidelines exactly.\n\nYour task is not to translate, paraphrase, normalise, or improve the input text. Your task is to annotate the tokens provided by the user.\n\nAlways preserve the input tokenisation. Do not add, remove, split, merge, reorder, transliterate, or translate tokens.\n\nWhen uncertain, use the authorised uncertainty mechanism rather than inventing an answer.\n\nReturn only the requested structured output. Do not include explanations outside the requested JSON.\n"}, {"role": "user", "content": "Annotate the following pre-tokenised sentence token by token.\n\nYou must follow the same annotation conventions as in the validated examples.\n\nLanguage: Ancient Greek\nScript: (unspecified)\nDomain: (unspecified)\nSentence ID: s2\n\nValidated examples:\n(none)\n\nNow annotate this sentence.\n\nOriginal text:\nμετοικήσαντες πόλιν, ἕως ἂν τὸ τοῦ χρόνου μῆκος ταύτην αὐτοῖς\n\nInput tokens:\n[\"μετοικήσαντες\", \"πόλιν,\", \"ἕως\", \"ἂν\", \"τὸ\", \"τοῦ\", \"χρόνου\", \"μῆκος\", \"ταύτην\", \"αὐτοῖς\"]\n\nTask:\nFor each input token, provide:\n1. the original surface form;\n2. a lemma;\n3. a UPOS tag;\n4. selected morphological features;\n5. a confidence value;\n6. an optional short local comment.\n\nAuthorised UPOS inventory:\n- `lemma`: free string (nullable)\n- `pos` ∈ {A, I+Adv, I+AdvPr, I+Conj, I+Intj, I+Neg, I+Part, I+Prep, LF, N+Ant, N+Com, N+Epi, N+Let, N+Pat, N+Prop, N+Top, NUM+Car, NUM+Ord, NUMA+Car, NUMA+Ord, PRO+Dem, PRO+Ind, PRO+Int, PRO+Per1p, PRO+Per1s, PRO+Per2p, PRO+Per2s, PRO+Pos1p, PRO+Pos1s, PRO+Pos2p, PRO+Pos2s, PRO+Rec, PRO+Ref, PRO+Rel, V} — or any combination of these tags joined by `@` (e.g. `TAG1@TAG2`).\n- `confidence` ∈ {low, medium, high}\n- `comment`: free string (nullable)\n\nAuthorised morphological features:\n(no morphological subfields)\n\nRules:\n- Return only valid JSON.\n- Do not translate the sentence.\n- Do not transliterate the tokens.\n- Do not normalise the surface forms.\n- Do not add, remove, split, merge, or reorder tokens.\n- The output must contain exactly one annotation per input token.\n- The value of `surface` must exactly match the corresponding input token.\n- Use only the authorised UPOS tags.\n- Use only the authorised feature names and values.\n- Follow the lemma conventions shown in the validated examples.\n- Follow the treatment of uncertain or ambiguous forms shown in the validated examples.\n- If uncertain, set `confidence` to `\"low\"` or `\"medium\"` and add a short local comment.\n- Keep `comment` short and local. Do not provide a full translation.\n\nExpected output schema:\n{\n  \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n  \"title\": \"HYE POS + lemma (bespoke tagset) annotation\",\n  \"type\": \"object\",\n  \"required\": [\n    \"sentence_id\",\n    \"language\",\n    \"tokens\"\n  ],\n  \"additionalProperties\": false,\n  \"properties\": {\n    \"sentence_id\": {\n      \"type\": \"string\"\n    },\n    \"language\": {\n      \"type\": \"string\"\n    },\n    \"tokens\": {\n      \"type\": \"array\",\n      \"items\": {\n        \"type\": \"object\",\n        \"required\": [\n          \"surface\",\n          \"lemma\",\n          \"pos\",\n          \"confidence\",\n          \"comment\"\n        ],\n        \"additionalProperties\": false,\n        \"properties\": {\n          \"surface\": {\n            \"type\": \"string\"\n          },\n          \"lemma\": {\n            \"type\": [\n              \"string\",\n              \"null\"\n            ]\n          },\n          \"pos\": {\n            \"type\": \"string\"\n          },\n          \"confidence\": {\n            \"type\": \"string\",\n            \"enum\": [\n              \"low\",\n              \"medium\",\n              \"high\"\n            ]\n          },\n          \"comment\": {\n            \"type\": [\n              \"string\",\n              \"null\"\n            ]\n          }\n        }\n      }\n    }\n  }\n}\n"}, {"role": "assistant", "content": "{\"sentence_id\": \"s2\", \"language\": \"Ancient Greek\", \"tokens\": [{\"surface\": \"μετοικήσαντες\"}, {\"surface\": \"πόλιν,\"}, {\"surface\": \"ἕως\"}, {\"surface\": \"ἂν\"}, {\"surface\": \"τὸ\"}, {\"surface\": \"τοῦ\"}, {\"surface\": \"χρόνου\"}, {\"surface\": \"μῆκος\"}, {\"surface\": \"ταύτην\"}, {\"surface\": \"αὐτοῖς\"}]}"}]}
{"messages": [{"role": "system", "content": "You are assisting with linguistic annotation for historical and under-resourced languages.\n\nYou must follow the annotation guidelines exactly.\n\nYour task is not to translate, paraphrase, normalise, or improve the input text. Your task is to annotate the tokens provided by the user.\n\nAlways preserve the input tokenisation. Do not add, remove, split, merge, reorder, transliterate, or translate tokens.\n\nWhen uncertain, use the authorised uncertainty mechanism rather than inventing an answer.\n\nReturn only the requested structured output. Do not include explanations outside the requested JSON.\n"}, {"role": "user", "content": "Annotate the following pre-tokenised sentence token by token.\n\nYou must follow the same annotation conventions as in the validated examples.\n\nLanguage: Ancient Greek\nScript: (unspecified)\nDomain: (unspecified)\nSentence ID: s2\n\nValidated examples:\n(none)\n\nNow annotate this sentence.\n\nOriginal text:\nμετοικήσαντες πόλιν, ἕως ἂν τὸ τοῦ χρόνου μῆκος ταύτην αὐτοῖς\n\nInput tokens:\n[\"μετοικήσαντες\", \"πόλιν,\", \"ἕως\", \"ἂν\", \"τὸ\", \"τοῦ\", \"χρόνου\", \"μῆκος\", \"ταύτην\", \"αὐτοῖς\"]\n\nTask:\nFor each input token, provide:\n1. the original surface form;\n2. a lemma;\n3. a UPOS tag;\n4. selected morphological features;\n5. a confidence value;\n6. an optional short local comment.\n\nAuthorised UPOS inventory:\n- `lemma`: free string (nullable)\n- `pos` ∈ {A, I+Adv, I+AdvPr, I+Conj, I+Intj, I+Neg, I+Part, I+Prep, LF, N+Ant, N+Com, N+Epi, N+Let, N+Pat, N+Prop, N+Top, NUM+Car, NUM+Ord, NUMA+Car, NUMA+Ord, PRO+Dem, PRO+Ind, PRO+Int, PRO+Per1p, PRO+Per1s, PRO+Per2p, PRO+Per2s, PRO+Pos1p, PRO+Pos1s, PRO+Pos2p, PRO+Pos2s, PRO+Rec, PRO+Ref, PRO+Rel, V} — or any combination of these tags joined by `@` (e.g. `TAG1@TAG2`).\n- `confidence` ∈ {low, medium, high}\n- `comment`: free string (nullable)\n\nAuthorised morphological features:\n(no morphological subfields)\n\nRules:\n- Return only valid JSON.\n- Do not translate the sentence.\n- Do not transliterate the tokens.\n- Do not normalise the surface forms.\n- Do not add, remove, split, merge, or reorder tokens.\n- The output must contain exactly one annotation per input token.\n- The value of `surface` must exactly match the corresponding input token.\n- Use only the authorised UPOS tags.\n- Use only the authorised feature names and values.\n- Follow the lemma conventions shown in the validated examples.\n- Follow the treatment of uncertain or ambiguous forms shown in the validated examples.\n- If uncertain, set `confidence` to `\"low\"` or `\"medium\"` and add a short local comment.\n- Keep `comment` short and local. Do not provide a full translation.\n\nExpected output schema:\n{\n  \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n  \"title\": \"HYE POS + lemma (bespoke tagset) annotation\",\n  \"type\": \"object\",\n  \"required\": [\n    \"sentence_id\",\n    \"language\",\n    \"tokens\"\n  ],\n  \"additionalProperties\": false,\n  \"properties\": {\n    \"sentence_id\": {\n      \"type\": \"string\"\n    },\n    \"language\": {\n      \"type\": \"string\"\n    },\n    \"tokens\": {\n      \"type\": \"array\",\n      \"items\": {\n        \"type\": \"object\",\n        \"required\": [\n          \"surface\",\n          \"lemma\",\n          \"pos\",\n          \"confidence\",\n          \"comment\"\n        ],\n        \"additionalProperties\": false,\n        \"properties\": {\n          \"surface\": {\n            \"type\": \"string\"\n          },\n          \"lemma\": {\n            \"type\": [\n              \"string\",\n              \"null\"\n            ]\n          },\n          \"pos\": {\n            \"type\": \"string\"\n          },\n          \"confidence\": {\n            \"type\": \"string\",\n            \"enum\": [\n              \"low\",\n              \"medium\",\n              \"high\"\n            ]\n          },\n          \"comment\": {\n            \"type\": [\n              \"string\",\n              \"null\"\n            ]\n          }\n        }\n      }\n    }\n  }\n}\n"}, {"role": "assistant", "content": "{\"sentence_id\": \"s3\", \"language\": \"Ancient Greek\", \"tokens\": [{\"surface\": \"ἂν\"}, {\"surface\": \"ἐνταῦθα\"}, {\"surface\": \"μακρῶν,\"}, {\"surface\": \"ἵνα\"}, {\"surface\": \"μηδὲν\"}, {\"surface\": \"τῶν\"}, {\"surface\": \"τῆς\"}, {\"surface\": \"πόλεως\"}, {\"surface\": \"τὴν\"}, {\"surface\": \"σὴν\"}]}"}]}