PDF Translation Tool Comparison

Google Translate vs DeepL vs doc2lang

Source Document: IMF Annual Report 2025 Financial Statements (Page 36)

Translation Direction: English → Japanese

Analysis Date: February 2025

1. Analysis Overview

This report compares the Japanese translations of the IMF Annual Report 2025 Financial Statements (English original) produced by three translation tools — Google Translate, DeepL (free version), and doc2lang — based on actual screenshot results.

The analysis focuses on Note page 36, which covers:

  • Sensitivity analysis of defined benefit obligations (table)
  • Other assets and liabilities (two tables)
  • Borrowings description (Section 13)

2. Visual Overview: Full-Page Comparison

Let us start by looking at the overall appearance of each translation result.

Original (English)

Original English PDF

Google Translate

Google Translate result

DeepL (Free Version)

DeepL translation result

doc2lang

doc2lang translation result

At first glance, doc2lang stays closest to the original layout, and DeepL also preserves the table structure well. Meanwhile, Google Translate inserts a watermark at the top of the page, and the DeepL free version adds a logo and promotional banner.

4. Sensitivity Analysis Table

This four-column table (assumption name, change range, increase impact, decrease impact) tests each tool's ability to handle tabular data.

Original

Sensitivity analysis original

Google Translate

Sensitivity analysis Google Translate

Issues found:

  1. Mistranslation in column header: The original "Assumption" is translated as "予測" (prediction) — a clear mistranslation; the correct term is "仮定" (assumption). Furthermore, the same word "assumption" is rendered as three different terms — "予測", "想定", and "仮定" — creating inconsistency within a single table
  2. Consistent notation: "0.5パーセント" (0.5 percent) is used uniformly across all rows (a positive point)
  3. "医療費の動向率": An acceptable translation of "Health care cost trend rate", though somewhat generic
  4. Table layout: Generally preserved; numerical relationships remain readable

DeepL

Sensitivity analysis DeepL

Issues found:

  1. Notable notation inconsistency: The most prominent issue is that the "Change in assumption" column mixes "0.5パーセント" and "0.5%" — the first row uses "0.5パーセント" while subsequent rows use "0.5%". This inconsistency within a single table is a significant issue for professional documents
  2. Mixed writing styles: The text immediately before the table ends with "〜は以下の通りです:" (polite style), while other sections use the formal "であった" style

doc2lang

Sensitivity analysis doc2lang

Key features:

  1. Consistent notation with "0.5 パーセント": The same format is used throughout every row, maintaining uniformity
  2. Column headers: Consistently translated as "仮定" (assumption), "仮定の変更" (change in assumption), "仮定の増加" (increase in assumption), "仮定の減少" (decrease in assumption)
  3. "数理仮定": "Actuarial assumptions" is translated as "数理仮定" — a concise and accurate technical term in insurance and pension terminology
  4. "感度分析": Same expression as DeepL's "感度分析" — Google Translate also uses "感度分析"

5. Other Assets Table

Original

Other assets original

Google Translate

Other assets Google Translate
  • Table structure is well preserved
  • "基本料金未収金" / "未収追加料金": "Basic charges receivable" and "Surcharges receivable" are translated using "料金" (fees) — too generic for the IMF context where "charges" carries specific meaning
  • "投資取引およびその他の受取手形": "Receivables" is translated as "受取手形" (notes receivable), which is a different accounting concept; "未収金" (receivables) would be more appropriate

DeepL

Other assets DeepL
  • Table structure is preserved. Styling changes (e.g., dotted borders) are present but the result remains readable
  • "基本料金未収入金" / "追加料金未収入金": "未収入金" is less standard than "未収金"
  • "投資取引及びその他の債権": "Receivables" is translated as "債権" — an appropriate accounting term
  • Unit notation: "(単位:百万SDR)" adds "単位:" (unit:) which is not in the original — a helpful clarification

doc2lang

Other assets doc2lang
  • Highest fidelity in table structure reproduction
  • "基本手数料未収金" / "追加手数料未収金": "Charges" is translated as "手数料" (commission/fees), which is closer to the financial context than "料金" (Google/DeepL) but slightly different from the official IMF term "賦課金". However, "手数料" is acceptable in general financial translation
  • "投資取引およびその他未収金": "未収金" (receivables) is the standard accounting term
  • Unit notation: "(SDR百万単位)"

6. Other Liabilities Table — Where Terminology Differences Are Most Pronounced

This table contains a concentration of IMF-specific terminology, making the differences between tools most apparent.

Original

Other liabilities original

Google Translate

Other liabilities Google Translate
OriginalGoogle Translate OutputIssue
Remuneration payable on members' reserve tranche positionRemuneration paid on 会員の準備金トランシェポジション"会員" (members) — in IMF context, "加盟国" (member countries) is the correct term, as "members" refers to sovereign nations
Refundable commitment fees on active arrangementsRefundable commitment 料金 on アクティブな契約"アクティブな" is too literal — "有効な" (effective) is more appropriate. "契約" (contract) — the IMF standard term is "取極" (arrangements)

DeepL

Other liabilities DeepL
OriginalDeepL OutputEvaluation
Remuneration payable on members' reserve tranche positionRemuneration paid on 加盟国の準備枠ポジション"加盟国" (member countries) is accurate. "準備枠" is an interpretive translation that works but is not the official IMF term
Refundable commitment fees on active arrangementsRefundable commitment fees on 有効な契約"有効な" (effective) is appropriate. "契約" (contract) — the IMF standard term is "取極" (arrangements)

doc2lang

Other liabilities doc2lang
Originaldoc2lang OutputEvaluation
Remuneration payable on members' reserve tranche positionRemuneration paid on 加盟国のリザーブ・トランシュ・ポジションBoth "加盟国" and "リザーブ・トランシュ・ポジション" closely match IMF terminology. Note: IMF documents also use "トランシェ"; the variation between "トランシュ" and "トランシェ" stems from different Japanese transliterations of the French word "tranche"
Refundable commitment fees on active arrangementsRefundable コミットメント・フィー on 有効な取極"取極" is the official IMF translation of "arrangements" and doc2lang is the only tool among the three to use it. "コミットメント・フィー" is also a commonly used katakana expression in financial practice

This is the key differentiator: Whether a tool translates "arrangements" as "取極" serves as a litmus test for IMF document translation expertise. Google Translate uses "契約" (contract), DeepL uses "契約" (and "取決め" in Section 13), while only doc2lang uses the correct "取極".

7. Section 13: Borrowings — Comprehensive Comparison of Style and Terminology

Original

Section 13 original

Google Translate

Section 13 Google Translate
  • Heading format collapse: "13.1 新規借入協定NAB は 40 の参加者による" — the heading "New Arrangements to Borrow" merges with the beginning of the body text, eliminating the distinction between heading and paragraph. This is a serious issue for practical use
  • Writing style: Uses polite "です/ます" form ("〜できます", "〜発効しました", "〜提供します"), which is somewhat informal for financial statement notes
  • Terminology: Uses "新規借入協定" and "信用協定" — the official IMF term is "取極"
  • "割当枠の財源": Translation of "quota resources". "クォータ資源" is more standard in IMF contexts
  • Punctuation issue: In the second paragraph, a new sentence begins immediately after a period without a space

DeepL

Section 13 DeepL
  • Clear headings: "13. 借入金" and "13.1 新たな借入取決め" are properly separated, with formatting close to the original
  • Highly natural Japanese: The sentence "GRAは、主にNABを主要な常設借入枠として、割当資源を一時的に補完するために借入を行うことができる" reads very fluently
  • Writing style: Consistent formal "である" style throughout ("〜できる", "〜がある", "〜であった")
  • Terminology: Uses "新たな借入取決め", which is close to but different from "取極". "執行理事会" is a literal translation of "Executive Board", but the official IMF Japanese term is "理事会"
  • Possible misreading of "全参加国": In the original, "all of which were effective" refers to arrangements, but DeepL appears to translate the subject as "all participating countries were effective" — potentially misidentifying the grammatical subject

doc2lang

Section 13 doc2lang
  • Headings: "13. 借入" and "13.1 新規借入取極" — uses the correct IMF term "取極"
  • "クォータ資源": "Quota resources" translated as "クォータ資源". Using the katakana "クォータ" for "quota" is standard practice in IMF documents
  • "信用取極の集合": Accurately deconstructs and translates "standing set of credit arrangements"
  • "IMFクォータに対する第二の防衛線": Accurate IMF terminology
  • Writing style: Consistent formal "である" style throughout
  • Line break issue: An unnatural line break occurs after "貸付契約および手形購入契約(二国間)", splitting "借入協定)から構成される。" onto the next line. In the original this is a single sentence, but the break placement is awkward

8. Key Terminology Cross-Reference

Original (English)Google TranslateDeepLdoc2langOfficial IMF Japanese
arrangements協定契約 / 取決め取極取極
charges (receivable)料金料金手数料賦課金
reserve tranche position準備金トランシェポジション準備枠ポジションリザーブ・トランシュ・ポジションリザーブ・トランシェ・ポジション
commitment feesコミットメント料金コミットメント料コミットメント・フィーコミットメント・フィー
quota resources割当枠の財源割当資源クォータ資源クォータ資源
Executive Board理事会執行理事会理事会理事会
active arrangementsアクティブな契約有効な契約有効な取極有効な取極
New Arrangements to Borrow新規借入協定新たな借入取決め新規借入取極新規借入取極
actuarial assumptions保険数理上の仮定保険数理上の前提条件数理仮定数理仮定
weighted average duration加重平均期間加重平均デュレーション加重平均デュレーション加重平均デュレーション
note purchase agreements債券購入契約債券購入契約手形購入契約手形購入契約
members会員加盟国加盟国加盟国
second line of defense第2の防衛線第二の防衛ライン第二の防衛線第二の防衛線

Bold text indicates the translation closest to the official IMF term among the three tools

Result: doc2lang used the closest (or matching) translation to the official IMF term in 11 out of 13 items. DeepL matched in 4 items, and Google Translate in 1 item.

9. Overall Evaluation Matrix

Evaluation CriteriaGoogle TranslateDeepLdoc2lang
Layout preservation (tables)
Layout preservation (headings)Collapsed at 13.1
Unwanted insertionsWatermark textLogo and ads (free version only)
IMF-specific terminology
General accounting termsreceivables → 受取手形 (incorrect)
Natural-sounding Japanese
Style consistencyMixed polite/formal stylesSome polite/formal style mixingConsistent formal style throughout
Notation consistencyMixed パーセント/% notation
Numerical accuracy
CostFreeSubscription-basedPaid
Ease of use

◎ = Excellent ○ = Good △ = Needs improvement

10. Strengths and Weaknesses of Each Tool

Google Translate

Strengths

  • Completely free, no registration required, instant access
  • Unmatched language coverage with 130+ languages
  • Table structure preservation has significantly improved in recent years
  • No errors in numerical data translation

Weaknesses

  • "Assumption" mistranslated as "予測" (prediction), with three different translations for the same word
  • Heading merges with body text in Section 13.1
  • Does not handle IMF-specific terminology
  • "Receivables" mistranslated as "受取手形" (notes receivable)
  • Polite "です/ます" style is too casual for financial statement notes

DeepL

Strengths

  • Most natural-sounding Japanese among the three tools
  • Relatively high accuracy in general accounting terminology
  • Good table structure preservation
  • Correctly uses terms like "加盟国" (member countries) in several places
  • Clear heading structure preservation

Weaknesses

  • Mixed "0.5パーセント" and "0.5%" notation within the same table
  • Inconsistent mixing of polite and formal writing styles
  • Does not use IMF-specific official terms like "取極" or "クォータ"
  • "Executive Board" translated as "執行理事会" instead of the IMF standard "理事会"
  • Free version includes logo and promotional ads

doc2lang

Strengths

  • Highest accuracy for IMF-specific terminology among all tools
  • Best layout reproduction fidelity
  • Consistent formal "である" writing style throughout
  • Uniform "0.5 パーセント" notation
  • No unwanted insertions whatsoever

Weaknesses

  • Paid service
  • Unnatural line break placement in Section 13
  • Writing style is slightly more rigid compared to DeepL
  • Translates "charges" as "手数料" (commission), which slightly differs from the official IMF term "賦課金"

11. Pricing Comparison

11.1 Pricing Plans for Each Tool

Google Translate

ItemDetails
Web VersionCompletely free, no registration required, no character limits
Cloud Translation API500,000 characters/month free. Overage: $20 per 1M characters
Supported File FormatsPDF, DOCX, PPTX, etc.
RemarksThe personal web version is completely free with unlimited use

DeepL

PlanMonthly PriceCharacter Limit/MonthFile Translation
FreeFree1,500 characters per translation1 file/month (non-editable)
Individual$10.49300,000 characters3 files/month (editable)
Team$34.49/user1,000,000 characters10 files/month (editable)
Business$68.99/userUnlimited100 files/month (editable)
EnterpriseContact salesCustomCustom
  • Annual billing saves approximately 33% compared to monthly billing
  • Free version includes logo/ads and has file size limitations
  • API Pro: $5.49/month (base) + $25 per 1M characters (usage-based)

doc2lang

Token CountUnit Price
1–20K tokens$0.50 / 1K tokens
20K–60K tokens$0.25 / 1K tokens
60K+ tokens$0.10 / 1K tokens
  • No subscription required — pay-per-use billing model
  • In English: 1 character ≈ 1 token, 1 page ≈ 2,000–3,000 tokens
  • Bulk credit purchases waive the minimum charge ($5), and credits never expire
  • Preview available before translation: no charge if you are not satisfied with the quality

11.2 Real-World Cost Simulation

Cost estimates for translating the IMF financial statements used in this analysis (1 page, approx. 3,000 characters):

ToolCostRemarks
Google Translate$0Completely free
DeepL Free$0Within free tier (but includes logo/ads)
DeepL Individual$10.49/月Requires monthly subscription (other translations included)
doc2langApprox. $1.503K tokens × $0.50/1K tokens

For a full 50-page report (approx. 150,000 characters):

ToolCostCalculation Basis
Google Translate$0Web version is completely free
DeepL FreeNot availableExceeds the 1,500 character per translation limit
DeepL Individual$10.49/月Within the monthly 300,000 character limit
DeepL Team$34.49/月For multi-user scenarios
doc2langApprox. $2920K×$0.50 + 40K×$0.25 + 90K×$0.10

11.3 Differences in Pricing Models

The three tools have fundamentally different pricing structures:

  • Google Translate: The web version's fully free model is an overwhelming advantage. Even the API offers 500,000 characters/month free of charge
  • DeepL: Subscription-based monthly billing. Cost-effectiveness improves with frequent use, making it ideal for regular users who translate often
  • doc2lang: Pay-per-use with no subscription required. Perfect for one-off needs like "just this month" or "just this document" — no cost incurred during months with no usage

12. Recommendations by Use Case

Use CaseRecommended ToolReason
Quick overview of contentGoogle TranslateFree and instant. Sufficient for grasping the general meaning
Internal reference material for team sharingDeepLNatural-sounding Japanese that reads well, at a reasonable cost
Submission to clients or external organizationsdoc2langScenarios requiring accuracy in specialized terminology and layout fidelity
Translation into multiple languages neededGoogle TranslateUnmatched language coverage
Regular translation of financial/international organization documentsdoc2langAdherence to domain-specific official terminology is essential
Balancing quality and costDeepLSuitable for the broadest range of users

13. Conclusion

Each of the three translation tools has distinct strengths, and the "best" tool depends on your specific use case.

Google Translate is free and instantly accessible, with significantly improved table structure preservation in recent years. While specialized terminology accuracy remains a challenge, it delivers sufficient quality for grasping the general content. It is the best choice for budget-constrained situations or casual information gathering.

DeepL produces the most fluent Japanese among the three tools and offers an excellent balance of quality and price. Despite some inconsistencies in notation, it is the most recommended tool for general business document translation. The paid version also eliminates ad insertions.

doc2lang handles specialized terminology unique to international organizations like the IMF — such as "取極", "クォータ", and "コミットメント・フィー" — with the highest accuracy, and surpasses the other two tools in layout fidelity and style consistency. It is the ideal choice for professional scenarios that demand formal, publication-ready translations.

Ultimately, the most practical approach is not to commit to a single tool, but to choose the right one based on your purpose, budget, and required quality level.

We also have a detailed guide on translating IMF financial statements with doc2lang.See the IMF translation guide

Try PDF Translation with doc2lang

Experience high-precision translations that handle even IMF-specific terminology