pymupdf · hashiromer · Dec 10, 2025 · Dec 10, 2025 · Dec 10, 2025 · Dec 10, 2025
diff --git a/README.md b/README.md
@@ -1,3 +1,38 @@
+# How to Run the Benchmark
+
+
+##  Installing dependencies
+``` bash
+cp .env.example .env # Add relevant API keys
+apt install tesseract-ocr 
+apt-get install -y libgl1 libglib2.0-0
+uv sync
+```
+
+Markdowns can be generated using the following command for each platform:
+``` bash
+uv run markdown_gen/{platform name}.py 
+```
+
+Docling is optimized for CPU with and without OCR, it will produce two folders `docling_ocr_results` (with OCR) and `docling_wocr_results` (without OCR). PyMuPDF-Layout is also optimized for CPU. 
+
+Running `uv run markdown_gen/{platform name}.py ` will produce a folder with this template 
+{platform_name}_results which will contain the markdowns folder containing all markdowns and a `results.jsonl` file containing the logs. 
+
+Docling and PyMuPDF_Layout additionally create a file inside the {platform_name}_results folder named `duration.txt` which contains the total time taken to generate the markdowns.
+
+## Running the benchmark
+
+``` bash
+uv run prod_benchmark.py
+```
+
+It produces these files
+
+1. benchmark_results_final.csv which contains the final results of the benchmark
+2. benchmark_granular.csv which contains the granular results of the benchmark for each eaxmple in the document
+
+
 # Flow-Aware PDF-to-Markdown Benchmark
 
 This repository provides a benchmark for evaluating the accuracy of PDF-to-Markdown extraction tools. The main goal is to measure how well a tool can convert a complex, 2D PDF document into a 1D (text/markdown) format that **preserves the logical reading flow** of the content.
@@ -44,19 +79,9 @@ We evaluate tools using a **Flow-Aware Text Accuracy (FATA) Score**.
 
 This benchmark was used to generate a comparative analysis of modern PDF extraction tools that produce markdown directly. The initial set of tools evaluated includes:
 
-* LlamaParse
-* Docling
-* Marker
-* Reducto
-* PyMuPDF4LLM
-* Pymupdf-Layout
-* Google Gemini (multimodal)
----
-
-
-# How to run this benchmark
-
-```bash
-uv sync
-uv run prod_benchmark.py
-```
+* [LlamaParse](https://www.llamaindex.ai/llamaparse) (Agentic Plus, premium)
+* [Docling](https://github.com/docling-project/docling) (Open Source, pipeline based with and without OCR)
+* [DataLab/Marker](https://www.datalab.to) (Hosted Solution)
+* [Reducto](https://reducto.ai) (Hosted Solution)
+* [PyMuPDF-Layout](https://pypi.org/project/pymupdf-layout/) (Available on PyPi with PolyForm Noncommercial License)
+* [Google Gemini 3](https://aistudio.google.com/models/gemini-3) (with a [single prompt](markdown_gen/ai_prompt.md))
diff --git a/benchmark_data/PDFs/.DS_Store b/benchmark_data/PDFs/.DS_Store
diff --git a/benchmark_data/PDFs/page_1.pdf b/benchmark_data/PDFs/page_1.pdf
diff --git a/benchmark_data/PDFs/page_10.pdf b/benchmark_data/PDFs/page_10.pdf
diff --git a/benchmark_data/PDFs/page_100.pdf b/benchmark_data/PDFs/page_100.pdf
diff --git a/benchmark_data/PDFs/page_101.pdf b/benchmark_data/PDFs/page_101.pdf
diff --git a/benchmark_data/PDFs/page_102.pdf b/benchmark_data/PDFs/page_102.pdf
diff --git a/benchmark_data/PDFs/page_103.pdf b/benchmark_data/PDFs/page_103.pdf
diff --git a/benchmark_data/PDFs/page_104.pdf b/benchmark_data/PDFs/page_104.pdf
diff --git a/benchmark_data/PDFs/page_105.pdf b/benchmark_data/PDFs/page_105.pdf
diff --git a/benchmark_data/PDFs/page_106.pdf b/benchmark_data/PDFs/page_106.pdf
diff --git a/benchmark_data/PDFs/page_107.pdf b/benchmark_data/PDFs/page_107.pdf
diff --git a/benchmark_data/PDFs/page_108.pdf b/benchmark_data/PDFs/page_108.pdf
diff --git a/benchmark_data/PDFs/page_109.pdf b/benchmark_data/PDFs/page_109.pdf
diff --git a/benchmark_data/PDFs/page_11.pdf b/benchmark_data/PDFs/page_11.pdf
diff --git a/benchmark_data/PDFs/page_110.pdf b/benchmark_data/PDFs/page_110.pdf
diff --git a/benchmark_data/PDFs/page_111.pdf b/benchmark_data/PDFs/page_111.pdf
diff --git a/benchmark_data/PDFs/page_112.pdf b/benchmark_data/PDFs/page_112.pdf
diff --git a/benchmark_data/PDFs/page_113.pdf b/benchmark_data/PDFs/page_113.pdf
diff --git a/benchmark_data/PDFs/page_114.pdf b/benchmark_data/PDFs/page_114.pdf
diff --git a/benchmark_data/PDFs/page_115.pdf b/benchmark_data/PDFs/page_115.pdf
diff --git a/benchmark_data/PDFs/page_116.pdf b/benchmark_data/PDFs/page_116.pdf
diff --git a/benchmark_data/PDFs/page_117.pdf b/benchmark_data/PDFs/page_117.pdf
diff --git a/benchmark_data/PDFs/page_118.pdf b/benchmark_data/PDFs/page_118.pdf
diff --git a/benchmark_data/PDFs/page_119.pdf b/benchmark_data/PDFs/page_119.pdf
diff --git a/benchmark_data/PDFs/page_12.pdf b/benchmark_data/PDFs/page_12.pdf
diff --git a/benchmark_data/PDFs/page_120.pdf b/benchmark_data/PDFs/page_120.pdf
diff --git a/benchmark_data/PDFs/page_121.pdf b/benchmark_data/PDFs/page_121.pdf
diff --git a/benchmark_data/PDFs/page_122.pdf b/benchmark_data/PDFs/page_122.pdf
diff --git a/benchmark_data/PDFs/page_123.pdf b/benchmark_data/PDFs/page_123.pdf
diff --git a/benchmark_data/PDFs/page_124.pdf b/benchmark_data/PDFs/page_124.pdf
diff --git a/benchmark_data/PDFs/page_125.pdf b/benchmark_data/PDFs/page_125.pdf
diff --git a/benchmark_data/PDFs/page_126.pdf b/benchmark_data/PDFs/page_126.pdf
diff --git a/benchmark_data/PDFs/page_127.pdf b/benchmark_data/PDFs/page_127.pdf
diff --git a/benchmark_data/PDFs/page_128.pdf b/benchmark_data/PDFs/page_128.pdf
diff --git a/benchmark_data/PDFs/page_129.pdf b/benchmark_data/PDFs/page_129.pdf
diff --git a/benchmark_data/PDFs/page_13.pdf b/benchmark_data/PDFs/page_13.pdf
diff --git a/benchmark_data/PDFs/page_14.pdf b/benchmark_data/PDFs/page_14.pdf
diff --git a/benchmark_data/PDFs/page_15.pdf b/benchmark_data/PDFs/page_15.pdf
diff --git a/benchmark_data/PDFs/page_16.pdf b/benchmark_data/PDFs/page_16.pdf
diff --git a/benchmark_data/PDFs/page_17.pdf b/benchmark_data/PDFs/page_17.pdf
diff --git a/benchmark_data/PDFs/page_18.pdf b/benchmark_data/PDFs/page_18.pdf
diff --git a/benchmark_data/PDFs/page_19.pdf b/benchmark_data/PDFs/page_19.pdf
diff --git a/benchmark_data/PDFs/page_2.pdf b/benchmark_data/PDFs/page_2.pdf
diff --git a/benchmark_data/PDFs/page_20.pdf b/benchmark_data/PDFs/page_20.pdf
diff --git a/benchmark_data/PDFs/page_21.pdf b/benchmark_data/PDFs/page_21.pdf
diff --git a/benchmark_data/PDFs/page_22.pdf b/benchmark_data/PDFs/page_22.pdf
diff --git a/benchmark_data/PDFs/page_23.pdf b/benchmark_data/PDFs/page_23.pdf
diff --git a/benchmark_data/PDFs/page_24.pdf b/benchmark_data/PDFs/page_24.pdf
diff --git a/benchmark_data/PDFs/page_25.pdf b/benchmark_data/PDFs/page_25.pdf
diff --git a/benchmark_data/PDFs/page_26.pdf b/benchmark_data/PDFs/page_26.pdf
diff --git a/benchmark_data/PDFs/page_27.pdf b/benchmark_data/PDFs/page_27.pdf
diff --git a/benchmark_data/PDFs/page_28.pdf b/benchmark_data/PDFs/page_28.pdf
diff --git a/benchmark_data/PDFs/page_29.pdf b/benchmark_data/PDFs/page_29.pdf
diff --git a/benchmark_data/PDFs/page_3.pdf b/benchmark_data/PDFs/page_3.pdf
diff --git a/benchmark_data/PDFs/page_30.pdf b/benchmark_data/PDFs/page_30.pdf
diff --git a/benchmark_data/PDFs/page_31.pdf b/benchmark_data/PDFs/page_31.pdf
diff --git a/benchmark_data/PDFs/page_32.pdf b/benchmark_data/PDFs/page_32.pdf
diff --git a/benchmark_data/PDFs/page_33.pdf b/benchmark_data/PDFs/page_33.pdf
diff --git a/benchmark_data/PDFs/page_34.pdf b/benchmark_data/PDFs/page_34.pdf
diff --git a/benchmark_data/PDFs/page_35.pdf b/benchmark_data/PDFs/page_35.pdf
diff --git a/benchmark_data/PDFs/page_36.pdf b/benchmark_data/PDFs/page_36.pdf
diff --git a/benchmark_data/PDFs/page_37.pdf b/benchmark_data/PDFs/page_37.pdf
diff --git a/benchmark_data/PDFs/page_38.pdf b/benchmark_data/PDFs/page_38.pdf
diff --git a/benchmark_data/PDFs/page_39.pdf b/benchmark_data/PDFs/page_39.pdf
diff --git a/benchmark_data/PDFs/page_4.pdf b/benchmark_data/PDFs/page_4.pdf
diff --git a/benchmark_data/PDFs/page_40.pdf b/benchmark_data/PDFs/page_40.pdf
diff --git a/benchmark_data/PDFs/page_41.pdf b/benchmark_data/PDFs/page_41.pdf
diff --git a/benchmark_data/PDFs/page_42.pdf b/benchmark_data/PDFs/page_42.pdf
diff --git a/benchmark_data/PDFs/page_43.pdf b/benchmark_data/PDFs/page_43.pdf
diff --git a/benchmark_data/PDFs/page_44.pdf b/benchmark_data/PDFs/page_44.pdf
diff --git a/benchmark_data/PDFs/page_45.pdf b/benchmark_data/PDFs/page_45.pdf
diff --git a/benchmark_data/PDFs/page_46.pdf b/benchmark_data/PDFs/page_46.pdf
diff --git a/benchmark_data/PDFs/page_47.pdf b/benchmark_data/PDFs/page_47.pdf
diff --git a/benchmark_data/PDFs/page_48.pdf b/benchmark_data/PDFs/page_48.pdf
diff --git a/benchmark_data/PDFs/page_49.pdf b/benchmark_data/PDFs/page_49.pdf
diff --git a/benchmark_data/PDFs/page_5.pdf b/benchmark_data/PDFs/page_5.pdf
diff --git a/benchmark_data/PDFs/page_50.pdf b/benchmark_data/PDFs/page_50.pdf
diff --git a/benchmark_data/PDFs/page_51.pdf b/benchmark_data/PDFs/page_51.pdf
diff --git a/benchmark_data/PDFs/page_52.pdf b/benchmark_data/PDFs/page_52.pdf
diff --git a/benchmark_data/PDFs/page_53.pdf b/benchmark_data/PDFs/page_53.pdf
diff --git a/benchmark_data/PDFs/page_54.pdf b/benchmark_data/PDFs/page_54.pdf
diff --git a/benchmark_data/PDFs/page_55.pdf b/benchmark_data/PDFs/page_55.pdf
diff --git a/benchmark_data/PDFs/page_56.pdf b/benchmark_data/PDFs/page_56.pdf
diff --git a/benchmark_data/PDFs/page_57.pdf b/benchmark_data/PDFs/page_57.pdf
diff --git a/benchmark_data/PDFs/page_58.pdf b/benchmark_data/PDFs/page_58.pdf
diff --git a/benchmark_data/PDFs/page_59.pdf b/benchmark_data/PDFs/page_59.pdf
diff --git a/benchmark_data/PDFs/page_6.pdf b/benchmark_data/PDFs/page_6.pdf
diff --git a/benchmark_data/PDFs/page_60.pdf b/benchmark_data/PDFs/page_60.pdf
diff --git a/benchmark_data/PDFs/page_61.pdf b/benchmark_data/PDFs/page_61.pdf
diff --git a/benchmark_data/PDFs/page_62.pdf b/benchmark_data/PDFs/page_62.pdf
diff --git a/benchmark_data/PDFs/page_63.pdf b/benchmark_data/PDFs/page_63.pdf
diff --git a/benchmark_data/PDFs/page_64.pdf b/benchmark_data/PDFs/page_64.pdf
diff --git a/benchmark_data/PDFs/page_65.pdf b/benchmark_data/PDFs/page_65.pdf
diff --git a/benchmark_data/PDFs/page_66.pdf b/benchmark_data/PDFs/page_66.pdf
diff --git a/benchmark_data/PDFs/page_67.pdf b/benchmark_data/PDFs/page_67.pdf
diff --git a/benchmark_data/PDFs/page_68.pdf b/benchmark_data/PDFs/page_68.pdf
diff --git a/benchmark_data/PDFs/page_69.pdf b/benchmark_data/PDFs/page_69.pdf
diff --git a/benchmark_data/PDFs/page_7.pdf b/benchmark_data/PDFs/page_7.pdf
diff --git a/benchmark_data/PDFs/page_70.pdf b/benchmark_data/PDFs/page_70.pdf
diff --git a/benchmark_data/PDFs/page_71.pdf b/benchmark_data/PDFs/page_71.pdf
diff --git a/benchmark_data/PDFs/page_72.pdf b/benchmark_data/PDFs/page_72.pdf
diff --git a/benchmark_data/PDFs/page_73.pdf b/benchmark_data/PDFs/page_73.pdf
diff --git a/benchmark_data/PDFs/page_74.pdf b/benchmark_data/PDFs/page_74.pdf
diff --git a/benchmark_data/PDFs/page_75.pdf b/benchmark_data/PDFs/page_75.pdf
diff --git a/benchmark_data/PDFs/page_76.pdf b/benchmark_data/PDFs/page_76.pdf
diff --git a/benchmark_data/PDFs/page_77.pdf b/benchmark_data/PDFs/page_77.pdf
diff --git a/benchmark_data/PDFs/page_78.pdf b/benchmark_data/PDFs/page_78.pdf
diff --git a/benchmark_data/PDFs/page_79.pdf b/benchmark_data/PDFs/page_79.pdf
diff --git a/benchmark_data/PDFs/page_8.pdf b/benchmark_data/PDFs/page_8.pdf
diff --git a/benchmark_data/PDFs/page_80.pdf b/benchmark_data/PDFs/page_80.pdf
diff --git a/benchmark_data/PDFs/page_81.pdf b/benchmark_data/PDFs/page_81.pdf
diff --git a/benchmark_data/PDFs/page_82.pdf b/benchmark_data/PDFs/page_82.pdf
diff --git a/benchmark_data/PDFs/page_83.pdf b/benchmark_data/PDFs/page_83.pdf
diff --git a/benchmark_data/PDFs/page_84.pdf b/benchmark_data/PDFs/page_84.pdf
diff --git a/benchmark_data/PDFs/page_85.pdf b/benchmark_data/PDFs/page_85.pdf
diff --git a/benchmark_data/PDFs/page_86.pdf b/benchmark_data/PDFs/page_86.pdf
diff --git a/benchmark_data/PDFs/page_87.pdf b/benchmark_data/PDFs/page_87.pdf
diff --git a/benchmark_data/PDFs/page_88.pdf b/benchmark_data/PDFs/page_88.pdf
diff --git a/benchmark_data/PDFs/page_89.pdf b/benchmark_data/PDFs/page_89.pdf
diff --git a/benchmark_data/PDFs/page_9.pdf b/benchmark_data/PDFs/page_9.pdf
diff --git a/benchmark_data/PDFs/page_90.pdf b/benchmark_data/PDFs/page_90.pdf
diff --git a/benchmark_data/PDFs/page_91.pdf b/benchmark_data/PDFs/page_91.pdf
diff --git a/benchmark_data/PDFs/page_92.pdf b/benchmark_data/PDFs/page_92.pdf
diff --git a/benchmark_data/PDFs/page_93.pdf b/benchmark_data/PDFs/page_93.pdf
diff --git a/benchmark_data/PDFs/page_94.pdf b/benchmark_data/PDFs/page_94.pdf
diff --git a/benchmark_data/PDFs/page_95.pdf b/benchmark_data/PDFs/page_95.pdf
diff --git a/benchmark_data/PDFs/page_96.pdf b/benchmark_data/PDFs/page_96.pdf
diff --git a/benchmark_data/PDFs/page_97.pdf b/benchmark_data/PDFs/page_97.pdf
diff --git a/benchmark_data/PDFs/page_98.pdf b/benchmark_data/PDFs/page_98.pdf
diff --git a/benchmark_data/PDFs/page_99.pdf b/benchmark_data/PDFs/page_99.pdf
diff --git a/benchmark_data/benchmark_filtered.csv b/benchmark_data/benchmark_filtered.csv
diff --git a/benchmark_filtered.csv b/benchmark_filtered.csv
diff --git a/benchmark_granular.csv b/benchmark_granular.csv
diff --git a/benchmark_results_final.csv b/benchmark_results_final.csv
@@ -1,8 +1,8 @@
 ,llama_parse,reducto,gemini,datalab,docling_cpu_with_ocr,pymupdflayout,docling_cpu_without_ocr,Folder_Count
-finance_samples,98.47,95.98,98.86,98.12,96.68,74.34,96.68,90.0
-govt tenders,96.14,91.88,96.0,93.01,93.54,92.82,93.54,26.0
-laws_sample,95.37,87.89,95.35,96.67,88.86,87.83,88.86,33.0
-manuals_sample,92.85,93.21,89.35,87.53,79.52,80.31,77.1,106.0
-patents,94.09,94.51,91.52,77.47,78.04,89.46,78.04,148.0
-science_sample,87.38,96.3,89.37,87.86,93.03,94.52,93.03,42.0
-Weighted_Mean,94.26,94.02,92.83,87.35,85.29,84.77,84.71,445.0
+finance_samples,98.47,95.98,98.86,98.18,96.68,74.34,96.68,90.0
+govt tenders,96.14,91.88,96.0,93.21,93.54,92.82,93.54,26.0
+laws_sample,95.24,87.89,95.35,96.67,88.86,87.83,88.86,33.0
+manuals_sample,92.43,93.21,89.35,87.53,79.52,80.31,77.1,106.0
+patents,94.17,94.51,91.52,77.91,78.04,89.46,78.04,148.0
+science_sample,87.16,96.3,89.37,87.86,93.03,94.52,93.03,42.0
+Weighted_Mean,94.16,94.02,92.83,87.53,85.29,84.77,84.71,445.0
diff --git a/datalab_results/markdowns/page_1.md b/datalab_results/markdowns/page_1.md
@@ -4,9 +4,9 @@ be formulated for the general SINR regime as follows:
 
 $$\begin{array}{rl}\underset{p_{n,k,l}}{\text{minimize}} & \log_2\prod_{l=1}^L\prod_{k=1}^K\prod_{n=1}^N\left(\frac{\sigma^2+I_{n,l}}{p_{n,k,l}h_{n,k,l}+\sigma^2+I_{n,l}}\right)^{\alpha_{n,k,l}} \\ \text{subject to} & \sum_{n=1}^N\alpha_{n,k,l}p_{n,k,l}\le P_{k,\max},\forall k,\forall l\end{array}\tag{12}$$
 
-Note that the numerator and denominator in (12) are posynomials and minimizing a ratio between two posynomials is referred to be a truly non-convex NP hard intractable problem known as complementary GP. However, this problem can be transformed into GP by letting the denominator  $f(p)=p_{n,k,l}h_{n,k,l}+\sigma^2+I_{n,l}=\sum_{l=1}^L\sum_{k=1}^Ku_{n,k,l}(p)$  and approximating the denominator  $f(p)$  with a monomial using the arithmetic/geometric mean inequality as follows:
+Note that the numerator and denominator in (12) are posynomials and minimizing a ratio between two posynomials is referred to be a truly non-convex NP hard intractable problem known as complementary GP. However, this problem can be transformed into GP by letting the denominator  $f(p)=p_{n,k,l}h_{n,k,l}+\sigma^2+I_{n,l}=\sum_{l=1}^L\sum_{k=1}^K u_{n,k,l}(p)$  and approximating the denominator  $f(p)$  with a monomial using the arithmetic/geometric mean inequality as follows:
 
-$$\sum_{l=1}^L\sum_{k=1}^Ku_{n,k,l}(p)\ge\prod_{l=1}^L\prod_{k=1}^K\left(\frac{u_{n,k,l}(p)}{s_{n,k,l}}\right)^{s_{n,k,l}}\tag{13}$$
+$$\sum_{l=1}^L\sum_{k=1}^K u_{n,k,l}(p)\ge\prod_{l=1}^L\prod_{k=1}^K\left(\frac{u_{n,k,l}(p)}{s_{n,k,l}}\right)^{s_{n,k,l}}\tag{13}$$
 
 where  $s_{n,k,l}=\frac{u_{n,k,l}(p_0)}{f(p_0)}$ . Thus, the problem can be solved by extending the single condensation method presented in [14] for multi-cell scenario. The details of centralized scheme A are presented in Algorithm 2.
 

diff --git a/datalab_results/markdowns/page_104.md b/datalab_results/markdowns/page_104.md
@@ -61,7 +61,7 @@ Bar chart showing Annualized Dividends Per Share from 2008 to 2013.
 
 Chart showing Adjusted Net Income and Adjusted Diluted Earnings Per Share from 2008 to 2013.
 
-Legend: Adjusted Net Income (\$ in Billions) (Gray bars), Adjusted Diluted Earnings Per Share (Line with circles).
+Legend: Adjusted Net Income (\$ in Billions) (Gray bars), Adjusted Diluted Earnings Per Share (Line with circles) (Red line).
 
 | Year | Adjusted Net Income (\$ in Billions) | Adjusted Diluted Earnings Per Share |
 |------|--------------------------------------|-------------------------------------|

diff --git a/datalab_results/markdowns/page_105.md b/datalab_results/markdowns/page_105.md
@@ -1,6 +1,6 @@
 
 
-## PROVEN PLUS PROBABLE RESERVES (SANTOS SHARE) BY ACTIVITY
+# PROVEN PLUS PROBABLE RESERVES (SANTOS SHARE) BY ACTIVITY
 
 |                                  | Sales gas<br>(incl. ethane)<br>PJ | Crude oil<br>mmbbl | Condensate<br>mmbbl | LPG<br>000<br>tonnes | Total<br>mmboe |
 |----------------------------------|-----------------------------------|--------------------|---------------------|----------------------|----------------|
@@ -11,7 +11,7 @@
 | Revisions                        | 126                               | -21                | 3                   | -45                  | 3              |
 | Estimated reserves year end 2004 | 2,873                             | 74                 | 49                  | 3,523                | 643            |
 
-## PROVEN PLUS PROBABLE RESERVES (SANTOS SHARE) YEAR END 2004 BY AREA (mmboe)
+# PROVEN PLUS PROBABLE RESERVES (SANTOS SHARE) YEAR END 2004 BY AREA (mmboe)
 
 | Area                        | Sales gas<br>(incl. ethane)<br>PJ | Crude oil<br>mmbbl | Condensate<br>mmbbl | LPG<br>000<br>tonnes | Total<br>mmboe |
 |-----------------------------|-----------------------------------|--------------------|---------------------|----------------------|----------------|
@@ -46,12 +46,8 @@ Geologists (AAPG). The definitions used are consistent with the requirements of
 
 Reserves are defined as those quantities of petroleum which are anticipated to be commercially recovered from known accumulations from a given date
 
-### forward. Santos reports reserves
+forward. Santos reports reserves net of the gas required for processing and transportation to the customer. Reserves reported are based on, and accurately reflect, information compiled by full-time employees of the Company who have the requisite qualifications and experience prescribed by the ASX Listing Rules.
 
-net of the gas required for processing and transportation to the customer. Reserves reported are based on, and accurately reflect, information compiled by full-time employees of the Company who have the requisite qualifications and experience prescribed by the ASX Listing Rules.
-
-## EXTERNALLY REVIEWED
-
-## BOOKING PROCESS
+## EXTERNALLY REVIEWED BOOKING PROCESS
 
 Santos' reserves processes and procedures were reviewed by independent expert, Gaffney, Cline & Associates, and found to be 'appropriate to providing robust estimates of Santos' reserve position in accordance with international industry practice'.
diff --git a/datalab_results/markdowns/page_107.md b/datalab_results/markdowns/page_107.md
@@ -16,60 +16,16 @@ To support continued innovation throughout the company, we made significant inve
 
 ![](0a070e8054da2df07205172e048f6659_img.jpg)
 
-Bar chart showing net sales (in Billions \$) from 2000 to 2004. The highest value is \$4.8 in 2004.
-
-| Year | Net Sales (Billions \$) |
-|------|-------------------------|
-| 00   | 3.2                     |
-| 01   | 3.5                     |
-| 02   | 3.5                     |
-| 03   | 3.8                     |
-| 04   | 4.8                     |
-
-net sales  
-(Billions \$) CAGR 8.3%
+Bar chart showing net sales (in Billions \$) from 2000 to 2004. The Y-axis ranges from 0 to 4. The sales figures are: 2000: 3.2, 2001: 3.5, 2002: 3.6, 2003: 3.8, 2004: 4.8. The chart is labeled "net sales (Billions \$) CAGR 8.3%".
 
 ![](88096eec96db1b919e141720a47f97c6_img.jpg)
 
-Bar chart showing net earnings (in Millions \$) from 2000 to 2004. The highest value is \$231.7 in 2004.
-
-| Year | Net Earnings (Millions \$) |
-|------|----------------------------|
-| 00   | 160                        |
-| 01   | 170                        |
-| 02   | 180                        |
-| 03   | 170                        |
-| 04   | 231.7                      |
-
-net earnings  
-(Millions \$) CAGR 8.0%
+Bar chart showing net earnings (in Millions \$) from 2000 to 2004. The Y-axis ranges from 0 to 200. The earnings figures are: 2000: 160, 2001: 170, 2002: 180, 2003: 175, 2004: 231.7. The chart is labeled "net earnings (Millions \$) CAGR 8.0%".
 
 ![](1066b22ea190b81c91c3441ea7618236_img.jpg)
 
-Bar chart showing net earnings per share (in Dollars) from 2000 to 2004. The highest value is \$1.65 in 2004.
-
-| Year | Net Earnings per Share (Dollars) |
-|------|----------------------------------|
-| 00   | 1.05                             |
-| 01   | 1.15                             |
-| 02   | 1.20                             |
-| 03   | 1.15                             |
-| 04   | 1.65                             |
-
-net earnings per share  
-(Dollars) CAGR 8.3%
+Bar chart showing net earnings per share (in Dollars) from 2000 to 2004. The Y-axis ranges from 0.00 to 1.40. The earnings per share figures are: 2000: 1.05, 2001: 1.15, 2002: 1.20, 2003: 1.15, 2004: 1.65. The chart is labeled "net earnings per share (Dollars) CAGR 8.3%".
 
 ![](59c76d920f3faf3933d80a47647d42a9_img.jpg)
 
-Bar chart showing annual dividend (in Dollars) from 2000 to 2004. The highest value is \$0.45 in 2004.
-
-| Year | Annual Dividend (Dollars) |
-|------|---------------------------|
-| 00   | 0.30                      |
-| 01   | 0.35                      |
-| 02   | 0.40                      |
-| 03   | 0.40                      |
-| 04   | 0.45                      |
-
-annual dividend  
-(Dollars) CAGR 6.5%
+Bar chart showing annual dividend (in Dollars) from 2000 to 2004. The Y-axis ranges from 0.00 to 0.40. The dividend figures are: 2000: 0.30, 2001: 0.35, 2002: 0.35, 2003: 0.40, 2004: 0.45. The chart is labeled "annual dividend (Dollars) CAGR 6.5%".
diff --git a/datalab_results/markdowns/page_109.md b/datalab_results/markdowns/page_109.md
@@ -6,8 +6,7 @@ Barcode image
 
 US009423690B2
 
-(12) **United States Patent**  
-**Takizawa et al.**
+# (12) **United States Patent** **Takizawa et al.**
 
 (10) **Patent No.:** **US 9,423,690 B2**  
 (45) **Date of Patent:** **Aug. 23, 2016**
@@ -34,7 +33,7 @@ US 2015/0168834 A1 Jun. 18, 2015
 (63) Continuation of application No.  
 PCT/JP2013/072485, filed on Aug. 16, 2013.
 
-## **Foreign Application Priority Data**
+### **Foreign Application Priority Data**
 
 Aug. 20, 2012 (JP) .................... 2012-181892  
 Mar. 15, 2013 (JP) .................... 2013-054401
@@ -75,7 +74,7 @@ See application file for complete search history.
 ### **U.S. PATENT DOCUMENTS**
 
 8,252,504 B2\* 8/2012 Harada et al. ........ 430/270.1  
-8,426,115 B2\* 4/2013 Hatakeyama et al. ....... 430/312
+8,426,115 B2\* 4/2013 Hatakeyama et al. ........ 430/312
 
 (Continued)
 
@@ -102,10 +101,10 @@ There is provided a pattern forming method, including: (a) forming a film by usi
 
 (1-0)
 
-$$\begin{array}{c} \text{*} \\ | \\ \text{R}_a \\ | \\ \text{L}_1 \\ | \\ \text{O} \\ | \\ \text{O} \\ | \\ \text{R}_1 \\ | \\ \text{R}_2 \\ | \\ \text{R}_{11} \\ | \\ \text{R}_{12} \\ | \\ \text{R}_{13} \end{array}$$
+$$\begin{array}{c} \text{R}_a \\ \text{C} \\ \text{L}_1 \\ \text{O} \\ \text{C} \\ \text{O} \\ \text{R}_1 \\ \text{R}_2 \\ \text{R}_{11} \\ \text{R}_{12} \\ \text{R}_{13} \end{array}$$
 
 (1-2)
 
-$$\begin{array}{c} \text{*} \\ | \\ \text{R}_b \\ | \\ \text{L}_2 \\ | \\ \text{C}_6\text{H}_4 \\ | \\ (\text{R}_3\text{R}_2) \\ | \\ (\text{OH})\text{R}_1 \end{array}$$
+$$\begin{array}{c} \text{R}_b \\ \text{C} \\ \text{L}_2 \\ \text{C}_6\text{H}_4 \\ (\text{R}_3\text{R}_2) \\ (\text{OH})\text{R}_1 \end{array}$$
 
 **14 Claims, No Drawings**
diff --git a/datalab_results/markdowns/page_111.md b/datalab_results/markdowns/page_111.md
@@ -12,7 +12,7 @@ The repeating unit having a cyclic carbonate ester structure is preferably the r
 
 $R_A^1$
 
-[Chemical structure of Formula (A-1): A repeating unit showing a carbonate group attached to a polymer backbone via  $R_A^1$ , linked to a cyclic carbonate ester structure (labeled A) via a divalent linking group Z, which is substituted with  $(R_A^2)_n$ .]
+[Chemical structure of Formula (A-1): A repeating unit showing a carbonate group attached to a polymer backbone via  $R_A^1$ , linked to a cyclic carbonate ester structure (labeled A) via a divalent group Z, which is substituted with  $(R_A^2)_n$ .]
 
 (A-1) 25
 
@@ -34,9 +34,9 @@ The substituent represented by  $R_A^2$  is, for example, an alkyl group, a cycl
 
 $n$  represents the number of substituents, and is an integer of 0 or more. For example,  $n$  is preferably 0 to 4, and more preferably 0.
 
-Examples of the divalent linking group represented by A may include an alkylene group, a cycloalkylene group, an ester bond, an amide bond, an ether bond, an urethane bond,
+Examples of the a divalent linking group represented by A may include an alkylene group, a cycloalkylene group, an ester bond, an amide bond, an ether bond, an urethane bond,
 
-an urea bond, or combination thereof. Examples of the alkylene group may include preferably an alkylene group having 1 to 10 carbon atoms, more preferably an alkylene group having 1 to 5 carbon atoms, and for example a methylene group, an ethylene group, a propylene group, and the like.
+an urea bond, or combination thereof. Examples of the alkylene group may include preferably an alkylene group having 1 to 10 carbon atoms, more preferably an alkylene group having 1 to 5 carbon atoms, and for example a methylene group, an ethylene group, an propylene group, and the like.
 
 In one aspect of the present invention, A is preferably a single bond, or an alkylene group.
 
@@ -46,7 +46,7 @@ The polycyclic group containing  $\text{O}—\text{C}(=\text{O})—\text{O}—$
 
 (a)
 
-[Chemical structure of Formula (a): A monocyclic carbonate ester structure, a 5-membered ring containing two oxygen atoms and one carbonyl group, with the carbonyl group double-bonded to one oxygen and single-bonded to the other oxygen, which is also bonded to the ring. The ring is labeled  $n_A$ .]
+[Chemical structure of Formula (a): A monocyclic carbonate ester structure, a 5-membered ring containing two oxygen atoms and a carbonyl group, with the ring size indicated by  $n_A$ .]
 
 The monomer corresponding to the repeating unit represented by Formula (A-1) may be synthesized by a conventionally known method described in, for example, Tetrahedron Letters, Vol. 27, No. 32 p. 3741(1986), Organic Letters, Vol. 4, No. 15 p. 2561(2002), and the like.
 
@@ -56,4 +56,4 @@ Hereinafter, specific examples of the repeating unit having a cyclic carbonate e
 
 In addition,  $R_A^1$  in the following specific examples has the same meaning as  $R_A^1$  in Formula (A-1).
 
-[Chemical structures showing two specific examples of repeating units having a cyclic carbonate ester structure. Both structures feature a carbonate group attached to a polymer backbone via  $R_A^1$ , linked to a cyclic carbonate ester structure (a 5-membered ring containing two oxygen atoms and one carbonyl group). The first example shows the cyclic carbonate ester attached directly to the backbone. The second example shows the cyclic carbonate ester attached via a divalent linking group.]
+[Chemical structures showing two specific examples of repeating units having a cyclic carbonate ester structure, both featuring a trifluoromethyl group ( $R_A^1$ ) attached to the polymer backbone.]