Document Translation with Command A Translate

1
#!pip install --upgrade cohere

1
# 1. Set up your Cohere client, translation prompt and maximum words per chunk 
2
import cohere
3
4
co = cohere.ClientV2("<YOUR API KEY>")
5
model = "command-a-translate-08-2025"
6
7
target_language = "Spanish"
8
prompt_template = "Translate everything that follows into {target_language}:\n\n"
9
max_words = 15  # Set your desired maximum number of words per chunk
10
11
# 2. Your source text
12
text = (
13
    "Enterprises rely on translation for some of their most sensitive and business-critical documents and cannot risk data leakage, compliance violations, or misunderstandings. Mistranslated documents can reduce trust and have strategic implications."
14
)
15
16
17
# 3. Define the chunk_split function (from earlier in your notebook)
18
def chunk_split(text, max_words, threshold=0.8):
19
20
    words = text.split()  # Turn the text into a list of words
21
    chunks = []  # Initialize an empty list to store our chunks
22
    start = 0  # Starting index for slicing the words list
23
24
    while start < len(words):
25
        # Determine the end index for the current chunk
26
        end = min(start + max_words, len(words))
27
        chunk_words = words[start:end]
28
        chunk_text = " ".join(chunk_words)  # Combine words back into a string
29
30
        # If we're at the end of the text or the chunk is too short, add it as is
31
        if end == len(words) or len(chunk_words) < max_words * threshold:
32
            chunks.append(chunk_text.strip())
33
            break
34
35
        # Try to find a natural breaking point within the chunk
36
        split_point = None
37
        for separator in ["\n", ".", ")", " "]:
38
            idx = chunk_text.rfind(separator)
39
            if idx != -1 and idx >= len(chunk_text) * threshold:
40
                split_point = idx + 1  # Position after the separator
41
                break
42
43
        if split_point:
44
            # If a good split point is found, add the chunk up to that point
45
            chunks.append(chunk_text[:split_point].strip())
46
            # Move the start index forward by the number of words consumed
47
            consumed = len(chunk_text[:split_point].split())
48
            start += consumed
49
        else:
50
            # If no good split point is found, add the entire chunk
51
            chunks.append(chunk_text.strip())
52
            start = end  # Move to the next chunk
53
54
    return chunks
55
56
# 4. Split the text into chunks using chunk_split
57
chunks = chunk_split(text, max_words=max_words)
58
59
# 5. Translate each chunk and collect results
60
translated_chunks = []
61
for chunk in chunks:
62
    prompt = prompt_template.format(target_language=target_language) + chunk
63
    response = co.chat(
64
        model=model,
65
        messages=[{"role": "user", "content": prompt}],
66
    )
67
    translated = response.message.content[0].text
68
    translated_chunks.append(translated)
69
70
# 6. Merge the translated chunks back together
71
translated_text = " ".join(translated_chunks)
72
73
# 7. Output the final translation
74
print(translated_text)